Research ArticleDimensional Evolution of Intelligent Cars Human-MachineInterface considering Take-Over Performance and DriversrsquoPerception on Urban Roads

Hao Yang12 Yueran Wang 3 and Ruoyu Jia1

1College of Mechanical and Material Engineering North China University of Technology Beijing China2Department of Industrial Engineering Tsinghua University Beijing China3School of International Art Education Tianjin Academy of Fine Arts Tianjin China

Correspondence should be addressed to Yueran Wang tafa_wyr163com

Received 11 July 2020 Revised 17 September 2020 Accepted 18 September 2020 Published 1 October 2020

Academic Editor Jun Yang

Copyright copy 2020 Hao Yang et al is is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

e study analyzed the driversrsquo take-over behaviors in intelligent cars when driving on urban roads and tried to find reasonabledimensions of the human-machine interface Firstly the main driving assistance functions in the process of take-over wereanalyzed based on the entropy theory and the weight values of each function for the consumerrsquos purchase intention werecalculated Secondly we explored the perceived comfortable dimensions of the interactive components under typical interactionmodes By means of experiments using a within-subjects design the initial population of the evolutionary computation wasobtainede evolutionary mechanism of dimensions driven by usersrsquo perception was constructed with a genetic algorithm Afterdebugging the parameters of the model we verified the rationality of the model and evolved appropriate dimensions Finally thevalidity of the evolved dimensions was proved by a controlled experiment and paired-sample t-test e results indicated that thecompletion time of most take-over tasks under the HMI with the evolved dimensions was significantly shorter which ensured theHMI could be more conducive to the take-over quality and traffic efficiency

1 Introduction

e intelligent vehicle is a comprehensive system whichintegrates environmental perception planning multileveldriving assistance and other functions Due to the com-plexity in intelligent vehicles brought about by technologiesin fields of computer modern sensing information fusioncommunication artificial intelligence and automatic con-trol a good human-vehicle interaction interface is importantin order to improve the safety and comfort of driving [1]With the growth of car ownership and the shortage of roadresources the complexity of urban traffic environment isalso increasing [2] e interaction performance and com-fort degree generated by the driving assistant system (DAS)would bring about driver pressure and distraction which isunfavorable to the sustainable development of traffic effi-ciency and road environment [3 4] According to the

automated driving standard J3016 [5] most of the currentautomated driving vehicles are at level 2-level 3 Because thislevel of intelligent system requires the driver to take over atany time the core of interaction design is human-vehiclecodrivingerefore the degree of technology acceptance byusers must be fully taken into consideration which is ofsignificance for a safe driving and sustainable urban traffic

Driving always puts a driver in a nervous state [3]erefore it is necessary to ensure that the driver can easilycontrol the steering wheel buttons combined switch pad-dles and other components while the body is not drasticallychanged is can help meet the physiological requirementsof drivers and reduce fatigue to ensure convenient rapidand effective operation is paper focuses on driversrsquo handextension problems in take-over tasks under human-com-puter codriving (HCCD) and looks for an optimal dimen-sion of the human-machine interface (HMI) based on

HindawiComplexityVolume 2020 Article ID 6519236 13 pageshttpsdoiorg10115520206519236

perceived comfortable dimensions (PCD) of the core in-teractive components

Driving movements are not completed independently byone part of the body but they are coordinated by multipleparts with coherence erefore the traditional staticmeasurement is difficult to obtain effective size data Whendesigning the dimension and layout of the HMI it is usefulto set the sizes based on the driverrsquos perception character-istics and comfort degree Because a userrsquos perception of aproduct performance may deviate due to the influence of itsappearance [6] this paper utilizes multiple measurements toobtain stable PCD data And the genetic algorithm (GA) wasused to evolve the optimal size

2 Literature Review

21 Cognitive and Behavioral Problems in the Process of Take-Over As a research hotspot in the field of transportationautomated driving can greatly improve driving safety andalleviate modern traffic problems such as accidents fuelemissions and road congestion [7ndash9] Meanwhile it canpromote driving comfort and reduce the impact of theenvironment [10] When the automated driving system isout of order or is unable to deal with the traffic conditionsthe driver needs to control the vehicle manually so thatdriving safety can be ensured e process to arouse thedriverrsquos cognition and behavioral control is called take-overerefore as a key link in the safe driving of conditionalautonomous vehicles the take-over process is a focus ofHCCD behavior research and interaction design

After receiving the take-over request the driver transfersthe attention to driving tasks then obtains the situationalawareness makes the decision and drives until the vehicle isfully controlled [11] If the size and distance of interactivecomponents such as buttons control levers or touch screensdo not meet the userrsquos expectations it is not conducive to thecompletion of take-over is is because when the systemissues a take-over request the driver who is performingnondriving tasks needs to quickly switch their attention tothe driving situation [11 12] In addition take-over tasksrequire the driver to respond in the shortest period It wasfound that the lead time affected the take-over quality[13ndash15] e interactive components that meet the userrsquospreferences in dimension and space can enable the user tocomplete the interactions quickly when taking over which isbeneficial to the improvement of take-over performance

Merat et al divided driving take-over into two partsoperation and cognition and found that the driver couldrecover the operational ability within 1-2 s but 6ndash10 s oreven more than 10 s were needed to recover the cognitiveability [16] is is related to the driverrsquos response and re-quires the design of buttons levers and other interactivecomponents should conform to the driverrsquos operatingcharacteristics and control force level When the controlright is switched the driverrsquos response refers to the ability toresume looking at the front in time and operating the vehicleafter receiving the take-over request from the system eresponse is mainly quantified by a variety of reaction timeand take-over time (from the time when the system sends

out take-over request to the time when the driver operatesthe steering wheel or pedals to achieve manual driving) [17]Gold et al quantified the reaction time of the driver whentaking over and the results showed that the average reactiontime of fixation putting the hand back to the steering wheeland looking around the rearview mirror was relatively 05 s15 s and 3 s [18] which provided a reference for the di-mension design of the interactive components

Making the design of the HMI accord with the userrsquosperceptual judgment is of great significance to a sustainableinteraction of human-computer codriving [19] e di-mension and spatial layout of an interface are related to thepreference of passengers [20]e size of the elements on theHMI determines the range of body motion and the usercomfort during the take-over process which is also closelyrelated to usability Usability is the suitability and efficiencyof user-interface interaction [21] In order to expand theusability of the HMI in the car the hierarchical interfaceproperties such as the size of the touch keys were studied[22 23] In addition in terms of take-over performance theaverage take-over reaction time of older drivers was at least12 s longer than that of younger drivers [24] while thedriving speed after taking over was lower than that of youngdrivers [25] erefore the efficiency problem caused by thedimension of the take-over interaction interface is veryimportant especially now that the aging phenomenon ismore and more serious Research showed that drivers fromdifferent age groups had significant differences in perceivedsafety and intention to use of automated driving and thescore increased with age [26]

22 Application of GA in Industrial Design and ErgonomicsIn industrial product design GA is often used in productgene research Based on the overall form of the product theshape is divided into several small feature units until itcannot be divided [27] At present related research studiesof product gene mostly adopt genetic operations such ascrossover and mutation to gradually cultivate a maturescheme after many iterations Chen et al obtained the geneexpression of the products based on the reverse deduction offunctional requirements [28] In the research of consumersrsquoperception and response to product appearance GA couldbe used to find the approximate optimal combination andappraise product appearance and color design [29] andevaluate peoplersquos emotion in the design process [30]

GA can solve the noise problem brought about bysubjective factors such as product usersrsquo cognition andpreference Take the noise source of interactive genetic al-gorithms (IGA) as an example Sun analyzed the relationshipbetween cognitive noise and evaluation accuracy [31] IGA isa branch of evolutionary computing (EC) which can evolveand generate design elements Kamalian et al put forwardthe idea of interval evaluation usersrsquo interaction behaviorsinterfered with the evolution process in the form of Paretorank shifting so that one evaluation was enough in the it-eration [32] Gu et al obtained the userrsquos perceptual eval-uation by the neural network so as to achieve fasterconvergence without increasing the userrsquos burden [33]

2 Complexity

In terms of the shape and size of HMI GA could de-termine the best material characteristics of the producthandle interface so as to achieve the optimal mechanicalresponse of the system by the fingertip model of the producthandle [34] In other fields of ergonomics GA was also apopular method In order to improve the efficiency andaccuracy of human shape prediction Cheng et al designed anew prediction method by using GA and BP neural network[35]e results showed that the prediction effect of this GA-BP model was better than that of BP SVM and K-meansmodels and it could accurately predict and cluster humanbody shapes In addition some research studies had applieda hybrid model based on GA classification tree and mul-tivariate adaptive expression splines to accurately predict themost influential risk factors for the upper limb musculo-skeletal diseases [36] Besides Sana et al proposed amathematical model of job scheduling problem for the er-gonomic factors of repetitive works lifting works andawkward postures in highly variable manufacturing envi-ronments [37] is was a multiobjective optimizationproblem with comprehensive ergonomic constraints whichwas solved by an improved nondominated sorting GA eresults improved the working conditions of 69 of theworkers GA could also be used to solve the optimizationdesign problem of walking worker manual assembly line[38] e goal of this study was to generate the requiredmodels at the lowest cost while maintaining an ergonomi-cally balanced operation ese studies all provided a basisfor applying GA to the evolution of human postures and size

3 Methods

31 Equipment and Procedure of the Experiment In thisstudy a driving simulator was used as the experimentplatform to collect the PCD task completion time and errortimes of the subjects in take-over tasks In order to modifythe controllers according to the optimal dimensions ob-tained by GA we chose a nonstandard simple drivingsimulator (Figure 1) In order to improve the applicabilityand validity of the data in urban environments the drivingscene of the simulator was set as city roads

New models of the side stalk and buttons conforming todimensions with the highest fitness would be made by 3Dprinters to replace the original ones en we rewrote thecodes to interconnect the printed models with the simulatorsystem By this means the task completion time and errortimes of the subjects under the new HMI could be collectedto analyze the differences in the interaction performancebetween the two HMIs A 3-spoke steering wheel with adiameter of Φ 365mm was used on the simulator Wedefined the function keys of the main take-over tasks on thesteering wheel and the gear lever panel e subjects weretrained to be familiar with the experimental process and thefunctions of each key

At present most mass-produced smart cars in the worldwere at L0 or L1 while others such as the latest Tesla Model-SX Volvo XC90 and Audi A8 were still between L2 and L3[17] erefore this paper analyzed the interaction modes ofthe self-adaptive cruise adjustment of the distance between

cars and auto hold which were core driving assistancefunctions (DAF) of automated driving between L1-L2 Andthe PCD reported by the subjects in the process of inter-action was collected

e length of cruise control rods in different vehicles wasdivergent from each other And for the cruise switch thatcontrolled by buttons the distance between the button andthe center of the driver or the steering wheel was also dif-ferent Due to the large diversity among the dimensions ofsteering wheels and interior space in multifarious vehicles itwas not significantly meaningful to analyze the absolute sizesof these controllers erefore PCD was collected in thisstudy rather than absolute sizesat is to take the horizontaland vertical centerlines of the steering wheel as referencesand collect the distance between the reference line and theperceived comfortable point on the controller In this waythe dimensions would be according to the userrsquos perceptualjudgment and preference e calibration points and rangesof the 14 dimensions involved in the three DAFs are shownin Table 1

e automatic transmission was the default option in theexperiment and speed was a controlled variable us thesubjects need not change gear ey were asked to completethe driving process of the whole virtual road sectionaccording to the prompts on the laptop screen which wasplaying interactive animation e prompts included thetask to be performed the number of times to perform acertain task and the components to be used in the next taske task flow of the experiment is shown in Figure 2

e specific procedures were as follows

(1) e subjects started the experiment in manualdriving modeWhen the system first prompted ldquoautoholdrdquo the subjects needed to press the A button onthe gear lever panel

M

D

S

E

B

A

C

Figure 1Wireframe of the driving simulator and the function keysused in the experiment

Complexity 3

(2) After stopping stably the subjects ought to startmanual driving again

(3) When the system first prompted that the self-adaptive cruise function could be used the subjectsturned the cruise control stalk D is was to sim-ulate switching the manual driving mode to theautomated driving mode and entering the state ofself-driving At this moment the virtual vehicle inthe system would run independently according tothe preset navigation route

(4) When the system prompted for adjusting the dis-tance between cars for the first time the subjectsneeded to press the M button

(5) en the subjects should complete auto hold self-adaptive cruise and distance adjustment in sequence

for the second time and pressed the buttons B E andS respectively

(6) Complete the auto hold operation for the third timeand press the C button

(7) After the completion of all the above tasks thesubjects were asked to point out the comfortableposition of the side stalk and buttons according tothe operating experience A researcher recorded thepoint with a marker pen on the simulator andcalibrated the sizese experiment used a within-subjects design erewere 13 subjects Each experiment lasted about 3minutes and the subjects needed to carry out theabove tasks five times to complete the experiment ofone day In order to ensure the accuracy and stability

Table 1 e driving assistance functions interaction modes and the analyzed dimensions

Driving assistancefunctions Interaction modes of take-over tasks Dimensions between the interactive component and the

steering wheel center

(A) Self-adaptive cruise

(A1) Switch the control right by the side stalk D

(a11) e PCD between the left endpoint of the stalk Dand the vertical centerline of the steering wheel (VCSW)(a12) e PCD between the left endpoint of the stalk Dand the horizontal centerline of the steering wheel(HCSW)

(A2) Switch the control right by the button E

(a21) e PCD between the vertical centerline (VC) ofthe button E and the VCSW(a22) e PCD between the horizontal centerline (HC)of the button E and the HCSW

(B) Adjustment of thedistance between cars

(B1) Adjust the distance by the button M which wasin the middle of the left spoke of the steering wheel

(b11)e PCD between the VC of the button M and theVCSW(b12)e PCD between the HC of the buttonM and theHCSW

(B2) Adjust the distance by the button S which wason the bottom edge of the left spoke of the steeringwheel

(b21) e PCD between the VC of the button S and theVCSW(b22) e PCD between the HC of the button S and theHCSW

(C) Auto hold

(C1) Start the function by button A which was underthe electric parking brake (EPB) button

(c11) e PCD between the VC of button A and theVCSW(c12) e PCD between the HC of button A and theHCSW

(C2) Start the function by button B which was on theleft of the EPB button

(c21) e PCD between the VC of button B and theVCSW(c22) e PCD between the HC of button B and theHCSW

(C3) Start the function by button C which was on theright of the EPB button

(c31) e PCD between the VC of button C and theVCSW(c32) e PCD between the HC of button C and theHCSW

0 1 2 3 4 5 6 7Time

Press thebutton C

Press thebutton S

Press thebutton E

Press thebutton B

Press thebutton M

Press thebutton A

Pull the sidestalk D

Start thesimulator in

manual drivingmode

None 1st auto hold 1st self-adaptivecruise

2nd self-adaptivecruise

1st adjustmentof the distancebetween cars

2nd adjustmentof the distancebetween cars

2nd auto hold 3rd auto hold

Figure 2 e task flow of the experiments

4 Complexity

of the perceptual experience the experiment wouldbe in progress for six days As a result each subjectwould carry out the tasks for 5lowast 6 30 times ePCD data collected each time were taken as the initialpopulation and the dimensions with the highestfitness would be obtained by GA

(8) In the 30th experiment the completion time anderror times of the subjects during the seven take-overtasks were collected

(9) After obtaining the PCD data with the highestadaptability models of the side stalk and buttonswere made by 3D printers to replace the originalones e subjects were asked to repeat the seventasks in Figure 2 again with the modified simulatorAnd their completion time and error times under thenew HMI would be collected

32 Information Entropy and Genetic Algorithm GA is aself-adaptive global optimization algorithm which imitatesthe natural selection and individual heredity of the biologicalworld [39] Traditionally interactive genetic algorithm(IGA) was widely used in product modelling design forsubject recognition [40] However IGA has some short-comings including the uncertainty of individual fitnessvalues the nonpersistence of the individual evaluationprocess and the nonuniqueness of optimization results [41]In this study GA was taken as the main research methodwhile PSC was used in the construction of fitness functionwhich integrated user preference intuition emotion andpsychological optimization into it erefore the fitnessvalues of evolutionary individuals could make the pop-ulation evolve in the direction that users expected

321 Fitness Function Whether the population can evolvein the direction of the ideal solution depends on the es-tablishment of fitness function to a great extent Accordingto the userrsquos PSC and the weight of the three DAFs thatneeded the driver to take over the artificial fitness functionwas formulated as follows

F 1

β1Q1 + β2Q2 + β3Q3( 1113857 (1)

Among which β1 sim β3 represented the importanceweights of the three DAFs which were obtained by calcu-lating the information entropy Q1 sim Q3 meant the com-fortable dimension of the operating components of eachDAF listed in Table 1 Q1 a11 lowast a12 + a21 lowast a22 Q2 b11 lowastb12 + b21 lowast b22 Q3 c11 lowast c12 + c21 lowast c22+ c31 lowast c32 emeanings of a11 sim c32 values are shown in Table 1 whichwere obtained in step (7) of the experiment For each in-teraction mode the comfortable position of the corre-sponding operation component was identified with a pointat first and then the distances between the point and thehorizontal and vertical centerlines of the steering wheel weremeasurede two distances were multiplied to get a relativecomfort area F was used to evaluate the fitness degree of

individuals With a smaller size the userrsquos behavioral path tocomplete the operations could be shorter and the taskcompletion time would be relatively less which were con-ducive to the completion of the take-over tasks ereforethe larger the F value was the more reasonable the sizeswere

322 Information Entropy and the Weight of the DrivingAssistance Functions Based on the entropy theory the studyexplored the importance level of the main DAFs in L1-L2automated driving and analyzed the influencing factors ofusersrsquo purchase intentions e uncertainty degree of thedivided sample set which was measured by means of cal-culating the information gain was used as the standard toweigh the quality of the division e larger the informationgain the less the uncertainty degree of the sample set

Entropy refers to the degree of system chaos which is ameasurement of the possibility of the system in a certainmacroscopic state Claude Elwood Shannon put forward theconcept of information entropy to express the order degreeof a system [42] Let S be a set of s samples Suppose theclassification attribute has m different values Ci(i

1 2 m) and let si be the number of samples in class Cien for a given sample its total entropy isI(s1 s2 sm) minus 1113936

mi1 Pilog2(Pi) Among which Pi is the

probability that any sample belongs to CiLet an attributeA have k different values a1 a2 ak1113864 1113865

Using the attribute to divide the set S into k subsetsS1 S2 Sk1113864 1113865 Among which Sj contains the samples with avalue aj in the set S Let sij be the number of samples of classCj in subset Sjen the information entropy of the samplesdivided according to A is given by

E(A) 1113944k

j1

s1j + s2j + middot middot middot + smj

sI s1j s2j smj1113872 1113873 (2)

where I(s1j s2j smj) minus 1113936mi1 Pijlog2(Pij) And Pij

((s1j + s2j + middot middot middot + smj)s) is the probability of the samples ofclass Cj in subset Sj

Finally the information gain obtained by dividing the setS according to the attribute A is Gain(A)

I(s1 s2 sm) minus E(A) Obviously the smaller the E(A)the larger the Gain(A) which means more information isprovided by A to judge the usersrsquo purchase intention and theimportance of A is higher After normalizing the Gain(A)

values of the three DAFs the weight values β1 sim β3 in thefitness function could be formed

We sent questionnaires online to eight professional usersto investigate their purchase decisions of products formed bydifferent combinations of interaction modes e purchasedecisions were divided into two categories which were ldquobuyrdquoand ldquonot buyrdquo Delphi method was used to collect theiropinions in an all-round way and ensure the consistency oftheir decisions Compared with taking the most frequentlyselected option as the final decision the results of the Delphimethod were more scientific and more information wasavailable in the research [43]

Complexity 5

323 Genetic Operation and Evolutionary Process In thissystem the premise to debug the genetic operation andmakejudgment was due to the irrationality and uncertainty of theshape and spatial dimension design of automobile con-trollers and panels the dimension evolution driven by usersrsquoexpectation was not only to search for the optimal solutionbut also to obtain the satisfactory solution under limitedresources erefore the model was different in codingselection crossover and mutation from traditional GAwhich was to solve problems represented by explicitfunctions

(1) Coding Generally there are two kinds of codingmethods for optimization problems which are real numbercoding and binary coding And both have advantages anddisadvantages Mapping errors exist in traditional binarycoding when continuous functions are discretizedWhen theindividual coding string is short the accuracy requirementsmay not be met While when the string is long although thecoding accuracy can be improved the search space of GAwill expand dramatically [44] erefore we used realnumber coding in this study

In the process of evolution a population Pop(t) wasformed Each chromosome Xt

i in the population representeddata from one of the 30 experiments e genes in thechromosome were composed of PCD reported by thesubjects after completing the whole experiment processwhich could be expressed as follows

Pop(t) Xt1 X

t2 X

ti X

tM1113966 1113967

Xti θt

i1 θti2 θt

ib θtiN1113960 1113961

(3)

where t represented the generation number of evolutions θrepresented gene coding and M represented the populationsize

(2) Initialized Population and Selection e populationneeded to be initialized before running GA According to theresearch demand PCD data from the 30 experiments weretaken as the primary population Each time data of the 14PCD indexes were collected to form the population matrix

e selection operation was the process of selectingindividuals from the previous generation of the populationto form the next generation e purpose of selection was toobtain fine individuals based on the fitness values so thatthey would have the opportunity to reproduce as parents forthe next generation Individuals with high fitness were morelikely to be inherited while the ones with low fitness wereless likely After debugging and comparison we chose themethod of normalized geometric select (NGS) to performthe selection operation NGS was mainly to sort the fitnessvalues and the better individuals would be maintained asparents is was more suitable for the selection of productdesign schemes and also helpful to prevent better individualsfrom being damaged

(3) Crossover and Mutation By crossover operation anew generation of individuals could be obtained whichcombined the characteristics of their parents and embodiedthe idea of information exchange In the case of real number

coding the arithmetic uniform crossover (AUC) betweenindividuals was generally utilized [45] AUC was a linearcombination of two individuals to produce two new onesemethod of crossover and getting the next generation wasas follows

c1 p1lowast a + p2lowast (1 minus a)

c2 p1lowast (1 minus a) + p2lowast a(4)

where p1 and p2 were the parents and a was a random mixamount

In this study the nonuniform mutation (NM) was usedto perform the mutation operation Let an individual beX X1X2 Xk Xl If Xk was the variation point andits value range was [Uk

min Ukmax] a new individual

X X1X2 Xkprime Xl could be obtained after non-

uniformly mutating at this point And the new gene valuewas given by

Xkprime

Xk + Δ t Ukmax minus Xk1113872 1113873 if random(0 1) 0

Xk minus Δ t Xk minus Ukmin1113872 1113873 if random(0 1) 1

⎧⎨

⎩

(5)

where Δ(t y) was a random number conforming to thenonuniform distribution in the range [0 y] y representedUk

max minus Xk and Xk minus Ukmin It was required that with the

increase of evolutionary generation number t the proba-bility of Δ(t y) approaching to 0 also raised gradually

In this study Δ(t y) was defined as follows

Δ(t y) y 1 minus r(1minus tT)b

1113872 1113873 (6)

where r was a random number conforming to the non-uniform distribution in the range [0 1] which was theselection pressure T was the maximum number of evolu-tionary generations in this study T 25 In the 25 iterationsthe better individuals with high fitness could be found out toexplore the evolutionary mechanism of the appropriate HMIdimensions of the DAFs b was a parameter to adjust thevariable step size which was a system parameter It deter-mined the dependence degree of the random number dis-turbance on t and the range was usually 2sim5 Afterdebugging the model and making comparisons repeatedlythis parameter was set as b 4 in this study

4 Results

41 Information Entropy and the Weights of the DAFs

4116e Expectation Information for the Classification of theHMI of Take-Over Tasks According to different combina-tions of the seven interaction modes in Table 1 a total of2lowast 2lowast 312 HMI schemes could be generated By means ofDelphi method purchase decisions for the 12 schemes wereinvestigated After four rounds of expert investigations thepurchase decisions of the eight professional users reached anagreement e classification results are shown in Table 2

According to the purchase intentions of the professionalusers 12 HMI schemes could be divided into two categories

6 Complexity

U1 1 (buy) and U2 2 (not buy) By summarizing andanalyzing the data the probabilities of the two categorieswere as follows P (U1) (512) and P (U2) (712) Basedon the formula the total entropy was as follows

I (U) minus512log2

512

minus712log2

712

09799 (7)

412 6e Conditional Entropy and Information Gain of EachInfluencing Factor In this study there were three factorsthat determined the purchase intentions of users A1 thestarting mode of the self-adaptive cruise A2 the position ofthe function key to adjust the distance between cars and A3the position of the function key of auto hold e proba-bilities of the factors and each condition are shown inTable 3

e information entropy values of the three factors wereas follows

E A1( 1113857 612

I(1 5) +612

I(4 2) 07842

E A2( 1113857 612

I(3 3) +612

I(2 4) 09592

E A3( 1113857 412

I(1 3) +412

I(1 3) +412

I(3 1) 08113

(8)

erefore the information gains of the factors were asfollows

Gain A1( 1113857 I(U) minus E A1( 1113857 01957

Gain A2( 1113857 I(U) minus E A2( 1113857 00207

Gain A3( 1113857 I(U) minus E A3( 1113857 01686

(9)

According to the information gains the starting mode ofthe self-adaptive cruise was the most important for the userrsquosperceived comfort and purchase intension while the posi-tion of the function key to adjust the distance between carswas not so significant By normalizing the information gainsof each influencing factor the results were taken as theweights of the factors which were

β1 Gain A1( 1113857

Gain A1( 1113857 + Gain A2( 1113857 + Gain A3( 1113857 05083

β2 Gain A2( 1113857

Gain A1( 1113857 + Gain A2( 1113857 + Gain A3( 1113857 00538

β3 Gain A3( 1113857

Gain A1( 1113857 + Gain A2( 1113857 + Gain A3( 1113857 04379

(10)

After taking these values into formula (1) the artificialfitness function could be constructed as follows

F 1

05083lowastQ1 + 00538lowastQ2 + 04379lowastQ3( 1113857 (11)

42 Results of GA

421 Characteristics of the Population e population sizeof this study was Pop 30 For the convenience of recordingthe unit of each dimension was decimeter (dm)e featuresof the 14 dimensions are described in Table 4

422 Solving Process of GA To match the userrsquos expecta-tion the seed of each initial population should be able to

Table 2 Classification of purchase intentions to interaction modes

e starting mode of self-adaptive cruise

e position of the function key to adjust thedistance between cars

e position of the function key ofauto-hold

Purchasedecision

Pulling the side stalk On the middle of the left spoke Under the EPB button 2Pressing the button On the middle of the left spoke Under the EPB button 2Pulling the side stalk On the bottom edge of the left spoke Under the EPB button 2Pressing the button On the bottom edge of the left spoke Under the EPB button 1Pulling the side stalk On the middle of the left spoke On the left of the EPB button 2Pressing the button On the middle of the left spoke On the left of the EPB button 1Pulling the side stalk On the bottom edge of the left spoke On the left of the EPB button 2Pressing the button On the bottom edge of the left spoke On the left of the EPB button 2Pulling the side stalk On the middle of the left spoke On the right of the EPB button 1Pressing the button On the middle of the left spoke On the right of the EPB button 1Pulling the side stalk On the bottom edge of the left spoke On the right of the EPB button 2Pressing the button On the bottom edge of the left spoke On the right of the EPB button 1

Table 3 e probabilities of the three factors and each condition

P (Vi) P (U1 | Vi) P (U2 | Vi)

A1Side stalk 612 16 56Button 612 46 26A2Middle 612 36 36Bottom edge 612 26 46A3Under the EPB button 412 14 34On the left of the EPB button 412 14 34On the right of the EPB button 412 34 14P (Vi) represents the probability of each condition of the influencingfactors P(U1 | Vi) and P(U2 | Vi) respectively indicate the conditionalprobabilities of the two categories

Complexity 7

cover part of the information of the PCD data And it shouldbe capable of satisfying the geometric description rules ofproduct evolution in terms of space and size [46] e initialpopulation of this paper came from repeated experiments inmany days and the perceptual experience and spatial at-tributes reflected by the data were stable e optimal so-lution of the model was obtained after 25 iterations Figure 3shows the evolutionary process of GA

In the figure the red line represents the best fitness valueof each generation And the blue dotted line describes themean fitness of each generation It could be seen from thefigure that the optimal value in the model increased with thenumber of iterations and converged into a straight line afterreaching the maximum value Until the optimal value of themodel remained unchanged the global optimal solutioncould be obtained

e genetic evolution stopped after 25 iterations It couldbe seen from the figure that although the deviation betweenthe optimal fitness values and the mean ones increased afterthe 9th iteration it continued to evolve which met theexpectation of evolution After the 21st iteration the devi-ation was close to 0

In the 25 generations of evolutionary iteration based onthe PCD data eight individuals with the highest fitness wereobtained which were the HMI dimension schemes that metthe usersrsquo expectation best in a certain generation of thepopulation It would be rational to make designs of HMIssuch as steering wheels and gear lever panels in the laterstage according to these fine values is could not onlyinherit the modelling genes of the HMIs but was alsocoupled to the user expectation to a certain extent e fi-nally generated dimensional data of the individuals with thehighest fitness are shown in Table 5

43 Significance of the Difference between the Original Di-mensions and the Evolved Ones To further validate whetherthe individuals with the highest fitness could bring about a

better take-over performance it was reasonable to collect theneeded time to complete driving assistant tasks under theHMI with the original sizes and the one with the evolvedsizes And the differences between the two sets of HMIdimensions could be analyzed

In the last of the 30 experiments we collected the time ofthe 13 subjects to complete the take-over tasks under theseven interaction modes in Table 1 Since the tasks had beenrepeated many times the subjects were skilled and the taskcompletion time would be relatively stable After making 3Dprinting models of the side stalk and the buttons accordingto the most suitable dimensions (Table 5) the drivingsimulator was reformeden the 13 subjects were invited torepeat the seven tasks with the new HMI and this set ofcompletion time was collected To judge whether the di-mensions could effectively improve the interaction perfor-mance and efficiency of driving information processingpaired-sample t-test was used on the two groups of data eresults are shown in Table 6

It could be seen from the results that except for A2 andC2 the completion time under the two sets of dimensions inthe process of the other five tasks all had a significant dif-ference (Figure 4)

Figure 4 shows that the completion time of each task wasshorter when the dimensions with high fitness were used onthe HMI is indicated that the driverrsquos information pro-cessing was faster which helped to improve the quality oftake-over and the ability to control the car As the driverrsquosattention to the HMI might compete with the execution ofdriving tasks for cognitive resources [47] and impair thedriving performance a shorter completion time of take-overtasks was conducive to the distribution of the driverrsquospsychological resources and more beneficial to trafficefficiency

It could be found by comparing the completion time ofeach task under the new HMI that (A) for starting the self-

Table 4 Features of the 14 dimensions

Tasks anddimensions

Mean(dm)

Stddeviation

Minimum(dm)

Maximum(dm)

A1 a11 19513 00789 18086 2091a12 06055 0063 05117 06977

A2 a21 08624 00242 08224 08979a22 00784 00477 00008 01487

B1 b11 13356 00633 12528 14475b12 00315 00169 00024 00588

B2 b21 13999 00526 13060 14952b22 0322 00139 03011 03497

C1 c11 29776 01254 28032 31940c12 3272 01414 30207 34928

C2 c21 27451 00305 27010 27966c22 32714 01304 30143 34819

C3 c31 32929 00596 32019 33963c32 3244 01523 30011 34932

2500735

0074

00745

0075

00755

0076

00765

0077

00775

0078

Fittn

ess

5 10 15 200Generation

Mean fitnessBest fitness

Figure 3 Solving process of GA

8 Complexity

adaptive cruise the completion time of operating with a sidestalk was shorter (B) for the task of adjusting the distancebetween cars the completion time would be shorter whenoperating with a button located in the middle of the steeringwheel spoke (C) for the task of auto hold a starting buttonon the left side of the EPB button could bring about a shortercompletion time ese results provided a reference for thefuture automobile interior design

5 Discussion

Driving assistant system (DAS) plays an important role inthe context of human-computer codriving As reportedDAS can help the driver to complete the intense drivingtasks effectively and abate the pressure level of the driver[48] erefore it is meaningful to make a study for thecontroller dimensions in the DAS e study collectedcomfortable dimensions perceived by the drivers in a relaxedstate e aim was to detect a set of reasonable dimensions

which could make drivers at a low level of stress In ad-dition it is valuable for drivers to adapt to the different roadenvironment and realize sustainable traffic to optimize thesizes of interactive components on the in-vehicle HMI emain goals of in-vehicle technologies and cooperativeservices were to reduce traffic congestion and improve roadsafety [4]

e study attempted to analyze the optimal HMI di-mensions suitable for take-over tasks e involved HMIcomponents such as the steering wheel self-adaptive cruiselever distance adjustment button and auto hold buttonwere all direct design elements for users to see and touchBecause the design and performance of a product wereimportant factors affecting the decision-making of con-sumers [49] we calculated the information entropy andinformation gains of the three kinds of take-over tasks forclassifying usersrsquo purchase intentions e informationtransmitted by a product was multivariant complex andfuzzy so that there would be dissipation in the process oftransmission erefore the image evaluation process ofproducts could be regarded as a chaotic state e feedbackinformation could be calculated by information entropy tocarry out a comprehensive evaluation e smaller the en-tropy of an image the clearer it was From the perspective ofchaos degree the 2nd factor in Table 3 (the position of thefunction key to adjust the distance between cars) broughtmore uncertainty to the purchase decision(E(A2) 09592) which indicated that little informationwas provided by the 2nd factor for judging the usersrsquo de-cisions On the contrary the uncertainty produced by the 1stfactor (the starting mode of the self-adaptive cruise) was thesmallest (E(A1) 07842) is illustrated that using theside stalk or a button to complete the switch of the controlright had a great impact on the satisfaction and purchasedecision of users

After normalizing the information entropy of threekinds of take-over tasks their weight values were calculated

Table 5 e finally generated dimensional data of the individuals with the highest fitness

(A) e starting mode of self-adaptive cruise

(B) e position of thefunction key to adjust thedistance between cars

(C) e position of the function key of auto-hold Fitness value

a11 a12 a21 a22 b11 b12 b21 b22 c11 c12 c21 c22 c31 c32202 062 089 009 136 004 148 030 284 302 271 338 321 313 008

Table 6 e paired sample t-test results of task completion time under the two HMIs with different dimensions

Interaction modes of take-over tasks

e task completiontime under the original

dimensions (s)

e task completiontime under the evolved

dimensions (s) t-value Sig

Mean Std dev Mean Std devA1lowast 0715 0185 0537 0167 2553 0025A2 0657 0181 0637 0186 1151 0272B1lowast 0648 013 0462 0206 2797 0016B2lowastlowast 0736 0106 0571 0124 3116 0009C1lowast 0753 0295 0562 0222 2515 0027C2 0667 0226 0522 0176 1999 0069C3lowast 069 0208 0537 0225 2355 0036

A1lowast B1lowast B2lowastlowast C1lowast C3lowast

0

02

04

06

08

1

12

Time under the original dimensionsTime under the evolved dimensions

Figure 4 e completion time of the tasks with significant dif-ferences under the two HMIs with different dimensions

Complexity 9

and the fitness function was constructed In this way thedimensions with the highest fitness could be found out bymeans of GA

e background and driving experience of the subjects weredifferent and their familiarity with the take-over tasks and theHMI was variant Besides the noise of usersrsquo subjective per-ception data was always high Taking the above factors intoconsideration we collected PCD data repeatedly from the same13 subjects to form the initial population From the evolutionarytrajectory it could be found that in the 8th 10th 11th 14th 17thand 21st iteration there appeared an optimal solution

GA was always used for parameter optimization bysearching the optimal solutions while seldom used in userresearch and environment perception analysis But in studiesof user-centred design it could search the set of satisfactorysolutions driven by user expectations [46] Many pieces ofresearch had utilized GA to improve the ergonomic effi-ciency or optimize interactive component sizes in an op-erating environment [34ndash38 50 51] In this study wereported a driver perception research which investigated themost comfortable HMI sizes under different interactionmodes of take-over tasks By means of normalized geometricselection arithmetic uniform crossover and nonuniformmutation GA was applied in the research of perceivedcomfortable dimensions and a set of reasonable dimensionswas obtained e dimensional evolutionary method dis-cussed in this study embodied the concept of dynamic in-teraction design which took the mutual influence andrestrictions among different dimensions into considerationCompared with traditional methods of measuring manytimes and taking the average this method was more ad-vanced And the study enriched theoretical researches in thefield of product design and collaborative optimization

Perceived comfort was a complex problem for its dif-ficulty to be quantified objectively But it was useful in er-gonomic explorations Researchers had found that theperceptual messages from product property and usage re-quirements were competent to the analysis of user experi-ence [52] erefore it was reasonable to probe suitabledimensions by perceived comfort For the study of HMIdesign effect in a specific environment these data couldproduce important and dependable results [53] Rationaldimensions of the automobile interior HMI played an im-portant role in the driverrsquos operation quality Caruso [54]studied driversrsquo experience on the seat by an experiment andfound that interior space influenced the driversrsquo ergonomicperformance [54] For the take-over tasks in the situation ofhuman-computer codriving the dimensions also affectedwhether the driver could safely achieve the control rightswitch in the shortest possible time

In this paper paired-sample t-test was used to testwhether the sizes with the highest fitness could bring about asignificant reduction in task completion time In the resultswe found that when the interface adopted the finally evolvedsizes the completion time of each task was shorter As thecompletion time of in-vehicle interactive tasks was related tothe vehicle speed and driver errors under the context ofhuman-computer codriving [21 55] a shorter completiontime would help the driver to control the speed better

6 Limitation

e study explored the reasonable dimensions of the hu-man-machine interface in intelligent cars Neverthelessthere were still some deficiencies in the equipment used inthis study e driving simulator was easier to control andreform and helped to reduce the subjectsrsquo pressure How-ever the immersion visual interactivity and fidelity degreewere not high enough After obtaining reliable conclusionsin this study outdoor experiment with a real vehicle could beconducted in the next stage

Besides the sample size of this study was limited Whenthe sample size was larger in future studies the represen-tativeness of the data would be better and the results of GAcould be further improved

Finally usersrsquo perception of the HMI dimensions de-viated from each other In the next phase of the studyelectroencephalography (EEG) signal in the process ofperforming the driving tasks could be used By this meansthe subjectsrsquo attention and intention when interacting withthe HMI could be reflected in an objective way [56 57] as asupplement to their subjective report

7 Conclusion

For HMI design of industrial products physical concretemodelling parameters such as the product shape size andlayout were the objective properties that users could directlyperceive and the basis for the product to be expressed Basedon the dynamic measurement principle this paper proposedan optimized design method for ergonomic dimensions byGA e results indicated that the method was effective andprovided a new way for product shape innovation andhuman-machine dimensional automatic design e mainconclusions of this study were as follows

(1) Based on the information entropy theory the weightvalues of the three driving assistance functions werecalculated e results showed that the informationgain of the starting mode of the self-adaptive cruisewas highest while the information gain of the po-sition of the function key to adjust the distancebetween cars was lowest which provided a theo-retical basis for the dimensional fitness evaluationmechanism of the HMI

(2) e evolutionary process of HMI dimensionsdriven by user expectation was studied and theevolution conditions were analyzed e studyconfirmed the feasibility of normalized geometricselection arithmetic uniform crossover and non-uniform mutation in iterative researches for hu-man-machine dimensions rough parameterdebugging this evolutionary method was proved tobe effective in obtaining appropriate sizes of HMIcomponents e model could make full use of theadvantages of GA to improve the user researchperformance such as few restrictions on the opti-mization high robustness and an easily achievedglobal search

10 Complexity

(3) After finding the dimensions with the highest fitnessand applying them to the design of the HMI within-subjects experiments showed that the completiontime of five take-over tasks was significantly shorterexcept for that of Task A2 and C2is indicated thatthe information processing of drivers under the HMIwith evolved dimensions was faster which was morebeneficial to the vehicle control and take-overquality Intelligent vehicles achieved intelligent in-formation exchange with people vehicles and roadsthrough in-vehicle sensor system and informationterminals erefore the evolved dimensions couldmake the vehicle apt to respond to the environmentand have more time to fully deal with the complextraffic environment and would help to improve thetraffic efficiency therefrom

Data Availability

e data used to support the findings of this study areavailable from the corresponding author upon request

Consent

e study respected and guaranteed the subjectsrsquo right toparticipate in the study and allowed them to withdraw from thestudy unconditionally at any stage e study strictly followedthe informed consent procedure e subjects were asked tosign a consent form to inform them of the purpose processduration of the experiment matters needing attention and themethods and purpose of collecting the data so as to reducetheir anxiety and confusion In the experiment the personalsafety and health of the subjects took priority According to theusage guidelines of the driving simulator we tried to avoid anypossible injury in driving tasks e privacy of the subjects wasstrictly protected e subjects were informed of the storageuse and confidentiality measures of their personal informationand image data in advance truthfully And we would notdisclose the information to a third party or spread them on theInternet without authorization

Conflicts of Interest

e authors declare that they have no conflicts of interest

Acknowledgments

is researchwas funded by BeijingUrbanGovernance ResearchProject (grant no 20XN244) Scientific Research Program ofBeijing Education Commission (grant no KM202010009003)Yuyou Talent Support Program of North China University ofTechnology (grant no 107051360018XN012018) Tianjin YouthTalents Reserve Support Program and Chinese ErgonomicsSociety amp Kingfar Joint Research Fund for Outstanding YoungScholars (grant number CES-Kingfar-2019-001)

References

[1] S Jin J Yang E Wang and J Liu ldquoe influence of high-speed rail on ice-snow tourism in north eastern ChinardquoTourism Management vol 78 Article ID 104070 2020

[2] J Yang R Yang J Sun T Huang and Q Ge ldquoe spatialdifferentiation of the suitability of ice-snow tourist destina-tions based on a comprehensive evaluation model in ChinardquoSustainability vol 9 no 5 Article ID 774 2017

[3] W-Y Chung T-W Chong and B-G Lee ldquoMethods to detectand reduce driver stress a reviewrdquo International Journal ofAutomotive Technology vol 20 no 5 pp 1051ndash1063 2019

[4] H Farah and H N Koutsopoulos ldquoDo cooperative systemsmake driversrsquo car-following behavior saferrdquo TransportationResearch Part C Emerging Technologies vol 41 pp 61ndash722014

[5] SAE Taxonomy and Definitions for Terms Related to DrivingAutomation Systems for On-Road Motor Vehicles SAEJ3016TM Warrendale PA USA 2018

[6] J Hoegg J W Alba and D W Dahl ldquoe good the bad andthe ugly influence of aesthetics on product feature judg-mentsrdquo Journal of Consumer Psychology vol 20 no 4pp 419ndash430 2010

[7] M Kyriakidis R Happee and J C F De Winter ldquoPublicopinion on automated driving results of an internationalquestionnaire among 5000 respondentsrdquo TransportationResearch Part F Traffic Psychology and Behaviour vol 32pp 127ndash140 2015

[8] F Meng and C Spence ldquoTactile warning signals for in-vehiclesystemsrdquo Accident Analysis amp Prevention vol 75 pp 333ndash346 2015

[9] J Wan and C Wu ldquoe effects of lead time of take-overrequest and nondriving tasks on taking-over control of au-tomated vehiclesrdquo IEEE Transactions on Human MachineSystems vol 48 no 6 pp 582ndash591 2018

[10] S Petermeijer P Hornberger I Ganotis J D Winter andK Bengler ldquoe design of a vibrotactile seat for conveyingtake-over requests in automated drivingrdquo in Proceedings of the8th International Conference on Applied Human Factors andErgonomics Los Angeles CA USA July 2017

[11] T Ito A Takata and K Oosawa ldquoTime required for take-overfrom automated to manual drivingrdquo in Proceedings of the SAE2016 World Congress and Exhibition Detroit MI USA April2016

[12] K Zeeb A Buchner and M Schrauf ldquoWhat determines thetake-over time An integrated model approach of driver take-over after automated drivingrdquo Accident Analysis amp Preven-tion vol 78 pp 212ndash221 2015

[13] B Mok M Johns K J Lee et al ldquoEmergency automation offunstructured transition timing for distracted drivers of au-tomated vehiclesrdquo in Proceedings of the IEEE InternationalConference on Intelligent Transportation Systems Las PalmasSpain September 2015

[14] S Samuel A Borowsky S Zilberstein and D L FisherldquoMinimum time to situation awareness in scenarios involvingtransfer of control from an automated driving suiterdquoTransportation Research Record Journal of the TransportationResearch Board vol 2602 no 2602 pp 115ndash120 2016

[15] W Vlakveld N van Nes J de Bruin L Vissers andM van der Kroft ldquoSituation awareness increases when drivershave more time to take over the wheel in a level 3 automated

Complexity 11

car a simulator studyrdquo Transportation Research Part F TrafficPsychology and Behaviour vol 58 pp 917ndash929 2018

[16] N Merat A H Jamson F C H Lai M Daly andO M J Carsten ldquoTransition to manual driver behaviourwhen resuming control from a highly automated vehiclerdquoTransportation Research Part F Traffic Psychology and Be-haviour vol 27 no 26 pp 274ndash282 2014

[17] C Z Wu H R Wu and N C Lyu ldquoReview of control switchand safety of human-computer driving intelligent vehiclerdquoJournal of Traffic and Transportation Engineering vol 18no 6 pp 131ndash141 2018

[18] C Gold D Dambock L Lorenz and K Bengler ldquoldquoTakeoverrdquo how long does it take to get the driver back into thelooprdquo Proceedings of the Human Factors and ErgonomicsSociety Annual Meeting vol 57 no 1 pp 1938ndash1942 2013

[19] H Yang Y Zhao and Y Wang ldquoIdentifying modeling formsof instrument panel system in intelligent shared cars a studyfor perceptual preference and in-vehicle behaviorsrdquo Envi-ronmental Science and Pollution Research vol 27 no 1pp 1009ndash1023 2020

[20] A Berkovich A Lu B Levine and A V Reddy ldquoObservedcustomer seating and standing behavior and seat preferenceson board subway cars in New York cityrdquo TransportationResearch Record Journal of the Transportation ResearchBoard vol 2353 no 1 pp 33ndash46 2013

[21] R Li Y V Chen C Sha and Z Lu ldquoEffects of interface layouton the usability of in-vehicle information systems and drivingsafetyrdquo Displays vol 49 pp 124ndash132 2017

[22] H Kim S Kwon J Heo H Lee and M K Chung ldquoe effectof touch-key size on the usability of in-vehicle informationsystems and driving safety during simulated drivingrdquo AppliedErgonomics vol 45 no 3 pp 379ndash388 2014

[23] H Kim and H Song ldquoEvaluation of the safety and usability oftouch gestures in operating in-vehicle information systemswith visual occlusionrdquo Applied Ergonomics vol 45 no 3pp 789ndash798 2014

[24] M Koerber C Gold D Lechner and K Bengler ldquoe in-fluence of age on the take-over of vehicle control in highlyautomated drivingrdquo Transportation Research Part F TrafficPsychology amp Behaviour vol 39 pp 19ndash32 2016

[25] H Clark A C Mclaughlin B Williams and J Feng ldquoPer-formance in takeover and characteristics of non-driving re-lated tasks during highly automated driving in younger andolder driversrdquo Proceedings of the Human Factors amp Ergo-nomics Society Annual Meeting vol 61 no 1 pp 37ndash41 2017

[26] I Lijarcio S A Useche J Llamazares and L MontoroldquoPerceived benefits and constraints in vehicle automationdata to assess the relationship between driverrsquos features andtheir attitudes towards autonomous vehiclesrdquo Data in Briefvol 27 Article ID 104662 2019

[27] K-Z Chen and X-A Feng ldquoVirtual genes of manufacturingproducts and their reforms for product innovative designrdquoProceedings of the Institution of Mechanical Engineers Part CJournal of Mechanical Engineering Science vol 218 no 5pp 557ndash574 2004

[28] K-Z Chen X-A Feng and X-C Chen ldquoReverse deductionof virtual chromosomes of manufactured products for theirgene-engineering-based innovative designrdquo Computer-AidedDesign vol 37 no 11 pp 1191ndash1203 2005

[29] H-C Tsai and J-R Chou ldquoAutomatic design support andimage evaluation of two-coloured products using colour as-sociation and colour harmony scales and genetic algorithmrdquoComputer-Aided Design vol 39 no 9 pp 818ndash828 2007

[30] S-W Hsiao C-F Hsu and K-W Tang ldquoA consultation andsimulation system for product color planning based on in-teractive genetic algorithmsrdquo Color Research amp Applicationvol 38 no 5 pp 375ndash390 2013

[31] Y Sun W-l Wang Y-w Zhao and X-j Liu ldquoUser imageoriented interactive genetic algorithm evaluation mode inproduct developmentrdquo Computer Integrated ManufacturingSystems vol 18 no 2 pp 276ndash281 2012

[32] R Kamalian E Yeh Z Ying A M Agogino and H TakagildquoReducing human fatigue in interactive evolutionary com-putation through fuzzy systems and machine learning sys-temsrdquo in Proceedings of the 2006 IEEE InternationalConference on Fuzzy Systems Vancouver Canada July 2006

[33] Z Gu M Xi Tang and J H Frazer ldquoCapturing aestheticintention during interactive evolutionrdquo Computer-AidedDesign vol 38 no 3 pp 224ndash237 2006

[34] G Harih M Borovinsek and Z Ren Optimisation ofProductrsquos Hand-Handle Interface Material Parameters forImproved Ergonomics Springer International PublishingCham Switzerland 2015

[35] P Cheng D Chen and J Wang ldquoClustering of the bodyshape of the adult male by using principal component analysisand genetic algorithm-BP neural networkrdquo Soft Computingvol 24 no 17 Article ID 13219 2020

[36] N M B Serrano P J G Nieto A S Sanchez F S Lasherasand P R Fernandez AHybrid Algorithm for the Assessment ofthe Influence of Risk Factors in the Development of Upper LimbMusculoskeletal Disorders Springer Cham Switzerland 2018

[37] S S Sankar O-M Holman A F Gazabon and C J AcevedoldquoApplication of genetic algorithm to job scheduling underergonomic constraints in manufacturing industryrdquo Journal ofAmbient Intelligence and Humanized Computing vol 10no 5 pp 2063ndash2090 2019

[38] A-Z Atiya L Lee and K Xing ldquoDeveloping a multi-ob-jective genetic optimisation approach for an operationaldesign of a manual mixed-model assembly line with walkingworkersrdquo Journal of Intelligent Manufacturing vol 27 no 5pp 1ndash17 2014

[39] P Jedlicka and T Ryba ldquoGenetic algorithm application inimage segmentationrdquo Pattern Recognition and Image Anal-ysis vol 26 no 3 pp 497ndash501 2016

[40] L V Campenhout J Frens K Overbeeke A Standaert andH Peremans ldquoPhysical interaction in a dematerializedworldrdquo International Journal of Design vol 7 no 1 pp 1ndash182013

[41] K Nandhini and S R Balasundaram ldquoImproving readabilitythrough individualized summary extraction using interactivegenetic algorithmrdquo Applied Artificial Intelligence vol 30no 7 pp 635ndash661 2016

[42] R M Gray Entropy and Information 6eory Springer NewYork NY USA 2nd edition 2011

[43] B Anderhofstadt and S Spinler ldquoFactors affecting the pur-chasing decision and operation of alternative fuel-poweredheavy-duty trucks in Germanymdasha Delphi studyrdquo Trans-portation Research Part D Transport and Environmentvol 73 pp 87ndash107 2019

[44] X-m Wu and Y Feng ldquoReactive power optimization ofdistribution network based on improved genetic algorithmrdquoJournal of Xirsquoan Shiyou University (Natural Science Edition)vol 30 no 3 pp 95ndash99 2015

[45] Z Ling andM-j Li ldquoState space evolutionary algorithm basedon non-uniform mutation operatorrdquo Computer Technologyand Development vol 28 no 9 pp 68ndash71 2018

12 Complexity

[46] W Hu and J Zhao ldquoAutomobile styling gene evolutiondriven by usersrsquo expectation imagerdquo Journal of MechanicalEngineering vol 47 no 16 pp 176ndash181 2011

[47] J J Scott and R Gray ldquoA comparison of tactile visual andauditory warnings for rear-end collision prevention in sim-ulated drivingrdquo Human Factors 6e Journal of the HumanFactors and Ergonomics Society vol 50 no 2 pp 264ndash2752008

[48] B Reimer B Mehler and J F Coughlin ldquoReductions in self-reported stress and anticipatory heart rate with the use of asemi-automated parallel parking systemrdquo Applied Ergonom-ics vol 52 pp 120ndash127 2016

[49] B DrsquoIppolito ldquoe importance of design for firms com-petitiveness a review of the literaturerdquo Technovation vol 34no 11 pp 716ndash730 2014

[50] S Asensio-Cuesta J A Diego-Mas L Canos-Daros andC Andres-Romano ldquoA genetic algorithm for the design of jobrotation schedules considering ergonomic and competencecriteriardquo International Journal of Advanced ManufacturingTechnology vol 60 no 9ndash12 pp 1161ndash1174 2012

[51] S-C Lee H-E Tseng C-C Chang and Y-M HuangldquoApplying interactive genetic algorithms to disassembly se-quence planningrdquo International Journal of Precision Engi-neering and Manufacturing vol 21 no 4 pp 663ndash679 2020

[52] M G Shishaev V V Dikovitsky and L V Lapochkina ldquoeexperience of building cognitive user interfaces of multido-main information systems based on the mental model ofusersrdquo in Proceedings of the 6th Computer Science On-LineConference (CSOC 2017) Prague Czech Republic April 2017

[53] S Lee H Alzoubi and S Kim ldquoe effect of interior designelements and lighting layouts on prospective occupantsrsquoperceptions of amenity and efficiency in living roomsrdquo Sus-tainability vol 9 no 7 Article ID 1119 2017

[54] G Caruso ldquoMixed reality system for ergonomic assessment ofdriverrsquos seatrdquo International Journal of Virtual Reality vol 10no 2 pp 69ndash79 2011

[55] M L Reyes and J D Lee ldquoEffects of cognitive load presenceand duration on driver eye movements and event detectionperformancerdquo Transportation Research Part F Traffic Psy-chology and Behaviour vol 11 no 6 pp 391ndash402 2008

[56] D Cisler P M Greenwood D M Roberts R McKendrickand C L Baldwin ldquoComparing the relative strengths of EEGand low-cost physiological devices in modeling attentionallocation in semiautonomous vehiclesrdquo Frontiers in HumanNeuroscience vol 13 Article ID 109 2019

[57] I-H Kim J-W Kim S Haufe and S-W Lee ldquoDetection ofbraking intention in diverse situations during simulateddriving based on EEG feature combinationrdquo Journal of NeuralEngineering vol 12 no 1 Article ID 016001 2014

Complexity 13

In terms of the shape and size of HMI GA could de-termine the best material characteristics of the producthandle interface so as to achieve the optimal mechanicalresponse of the system by the fingertip model of the producthandle [34] In other fields of ergonomics GA was also apopular method In order to improve the efficiency andaccuracy of human shape prediction Cheng et al designed anew prediction method by using GA and BP neural network[35]e results showed that the prediction effect of this GA-BP model was better than that of BP SVM and K-meansmodels and it could accurately predict and cluster humanbody shapes In addition some research studies had applieda hybrid model based on GA classification tree and mul-tivariate adaptive expression splines to accurately predict themost influential risk factors for the upper limb musculo-skeletal diseases [36] Besides Sana et al proposed amathematical model of job scheduling problem for the er-gonomic factors of repetitive works lifting works andawkward postures in highly variable manufacturing envi-ronments [37] is was a multiobjective optimizationproblem with comprehensive ergonomic constraints whichwas solved by an improved nondominated sorting GA eresults improved the working conditions of 69 of theworkers GA could also be used to solve the optimizationdesign problem of walking worker manual assembly line[38] e goal of this study was to generate the requiredmodels at the lowest cost while maintaining an ergonomi-cally balanced operation ese studies all provided a basisfor applying GA to the evolution of human postures and size

3 Methods

31 Equipment and Procedure of the Experiment In thisstudy a driving simulator was used as the experimentplatform to collect the PCD task completion time and errortimes of the subjects in take-over tasks In order to modifythe controllers according to the optimal dimensions ob-tained by GA we chose a nonstandard simple drivingsimulator (Figure 1) In order to improve the applicabilityand validity of the data in urban environments the drivingscene of the simulator was set as city roads

New models of the side stalk and buttons conforming todimensions with the highest fitness would be made by 3Dprinters to replace the original ones en we rewrote thecodes to interconnect the printed models with the simulatorsystem By this means the task completion time and errortimes of the subjects under the new HMI could be collectedto analyze the differences in the interaction performancebetween the two HMIs A 3-spoke steering wheel with adiameter of Φ 365mm was used on the simulator Wedefined the function keys of the main take-over tasks on thesteering wheel and the gear lever panel e subjects weretrained to be familiar with the experimental process and thefunctions of each key

At present most mass-produced smart cars in the worldwere at L0 or L1 while others such as the latest Tesla Model-SX Volvo XC90 and Audi A8 were still between L2 and L3[17] erefore this paper analyzed the interaction modes ofthe self-adaptive cruise adjustment of the distance between

cars and auto hold which were core driving assistancefunctions (DAF) of automated driving between L1-L2 Andthe PCD reported by the subjects in the process of inter-action was collected

e length of cruise control rods in different vehicles wasdivergent from each other And for the cruise switch thatcontrolled by buttons the distance between the button andthe center of the driver or the steering wheel was also dif-ferent Due to the large diversity among the dimensions ofsteering wheels and interior space in multifarious vehicles itwas not significantly meaningful to analyze the absolute sizesof these controllers erefore PCD was collected in thisstudy rather than absolute sizesat is to take the horizontaland vertical centerlines of the steering wheel as referencesand collect the distance between the reference line and theperceived comfortable point on the controller In this waythe dimensions would be according to the userrsquos perceptualjudgment and preference e calibration points and rangesof the 14 dimensions involved in the three DAFs are shownin Table 1

e automatic transmission was the default option in theexperiment and speed was a controlled variable us thesubjects need not change gear ey were asked to completethe driving process of the whole virtual road sectionaccording to the prompts on the laptop screen which wasplaying interactive animation e prompts included thetask to be performed the number of times to perform acertain task and the components to be used in the next taske task flow of the experiment is shown in Figure 2

e specific procedures were as follows

(1) e subjects started the experiment in manualdriving modeWhen the system first prompted ldquoautoholdrdquo the subjects needed to press the A button onthe gear lever panel

M

D

S

E

B

A

C

Figure 1Wireframe of the driving simulator and the function keysused in the experiment

Complexity 3

(2) After stopping stably the subjects ought to startmanual driving again

(3) When the system first prompted that the self-adaptive cruise function could be used the subjectsturned the cruise control stalk D is was to sim-ulate switching the manual driving mode to theautomated driving mode and entering the state ofself-driving At this moment the virtual vehicle inthe system would run independently according tothe preset navigation route

(4) When the system prompted for adjusting the dis-tance between cars for the first time the subjectsneeded to press the M button

(5) en the subjects should complete auto hold self-adaptive cruise and distance adjustment in sequence

for the second time and pressed the buttons B E andS respectively

(6) Complete the auto hold operation for the third timeand press the C button

(7) After the completion of all the above tasks thesubjects were asked to point out the comfortableposition of the side stalk and buttons according tothe operating experience A researcher recorded thepoint with a marker pen on the simulator andcalibrated the sizese experiment used a within-subjects design erewere 13 subjects Each experiment lasted about 3minutes and the subjects needed to carry out theabove tasks five times to complete the experiment ofone day In order to ensure the accuracy and stability

Table 1 e driving assistance functions interaction modes and the analyzed dimensions

Driving assistancefunctions Interaction modes of take-over tasks Dimensions between the interactive component and the

steering wheel center

(A) Self-adaptive cruise

(A1) Switch the control right by the side stalk D

(a11) e PCD between the left endpoint of the stalk Dand the vertical centerline of the steering wheel (VCSW)(a12) e PCD between the left endpoint of the stalk Dand the horizontal centerline of the steering wheel(HCSW)

(A2) Switch the control right by the button E

(a21) e PCD between the vertical centerline (VC) ofthe button E and the VCSW(a22) e PCD between the horizontal centerline (HC)of the button E and the HCSW

(B) Adjustment of thedistance between cars

(B1) Adjust the distance by the button M which wasin the middle of the left spoke of the steering wheel

(b11)e PCD between the VC of the button M and theVCSW(b12)e PCD between the HC of the buttonM and theHCSW

(B2) Adjust the distance by the button S which wason the bottom edge of the left spoke of the steeringwheel

(b21) e PCD between the VC of the button S and theVCSW(b22) e PCD between the HC of the button S and theHCSW

(C) Auto hold

(C1) Start the function by button A which was underthe electric parking brake (EPB) button

(c11) e PCD between the VC of button A and theVCSW(c12) e PCD between the HC of button A and theHCSW

(C2) Start the function by button B which was on theleft of the EPB button

(c21) e PCD between the VC of button B and theVCSW(c22) e PCD between the HC of button B and theHCSW

(C3) Start the function by button C which was on theright of the EPB button

(c31) e PCD between the VC of button C and theVCSW(c32) e PCD between the HC of button C and theHCSW

0 1 2 3 4 5 6 7Time

Press thebutton C

Press thebutton S

Press thebutton E

Press thebutton B

Press thebutton M

Press thebutton A

Pull the sidestalk D

Start thesimulator in

manual drivingmode

None 1st auto hold 1st self-adaptivecruise

2nd self-adaptivecruise

1st adjustmentof the distancebetween cars

2nd adjustmentof the distancebetween cars

2nd auto hold 3rd auto hold

Figure 2 e task flow of the experiments

4 Complexity

of the perceptual experience the experiment wouldbe in progress for six days As a result each subjectwould carry out the tasks for 5lowast 6 30 times ePCD data collected each time were taken as the initialpopulation and the dimensions with the highestfitness would be obtained by GA

(8) In the 30th experiment the completion time anderror times of the subjects during the seven take-overtasks were collected

(9) After obtaining the PCD data with the highestadaptability models of the side stalk and buttonswere made by 3D printers to replace the originalones e subjects were asked to repeat the seventasks in Figure 2 again with the modified simulatorAnd their completion time and error times under thenew HMI would be collected

32 Information Entropy and Genetic Algorithm GA is aself-adaptive global optimization algorithm which imitatesthe natural selection and individual heredity of the biologicalworld [39] Traditionally interactive genetic algorithm(IGA) was widely used in product modelling design forsubject recognition [40] However IGA has some short-comings including the uncertainty of individual fitnessvalues the nonpersistence of the individual evaluationprocess and the nonuniqueness of optimization results [41]In this study GA was taken as the main research methodwhile PSC was used in the construction of fitness functionwhich integrated user preference intuition emotion andpsychological optimization into it erefore the fitnessvalues of evolutionary individuals could make the pop-ulation evolve in the direction that users expected

321 Fitness Function Whether the population can evolvein the direction of the ideal solution depends on the es-tablishment of fitness function to a great extent Accordingto the userrsquos PSC and the weight of the three DAFs thatneeded the driver to take over the artificial fitness functionwas formulated as follows

F 1

β1Q1 + β2Q2 + β3Q3( 1113857 (1)

Among which β1 sim β3 represented the importanceweights of the three DAFs which were obtained by calcu-lating the information entropy Q1 sim Q3 meant the com-fortable dimension of the operating components of eachDAF listed in Table 1 Q1 a11 lowast a12 + a21 lowast a22 Q2 b11 lowastb12 + b21 lowast b22 Q3 c11 lowast c12 + c21 lowast c22+ c31 lowast c32 emeanings of a11 sim c32 values are shown in Table 1 whichwere obtained in step (7) of the experiment For each in-teraction mode the comfortable position of the corre-sponding operation component was identified with a pointat first and then the distances between the point and thehorizontal and vertical centerlines of the steering wheel weremeasurede two distances were multiplied to get a relativecomfort area F was used to evaluate the fitness degree of

individuals With a smaller size the userrsquos behavioral path tocomplete the operations could be shorter and the taskcompletion time would be relatively less which were con-ducive to the completion of the take-over tasks ereforethe larger the F value was the more reasonable the sizeswere

322 Information Entropy and the Weight of the DrivingAssistance Functions Based on the entropy theory the studyexplored the importance level of the main DAFs in L1-L2automated driving and analyzed the influencing factors ofusersrsquo purchase intentions e uncertainty degree of thedivided sample set which was measured by means of cal-culating the information gain was used as the standard toweigh the quality of the division e larger the informationgain the less the uncertainty degree of the sample set

Entropy refers to the degree of system chaos which is ameasurement of the possibility of the system in a certainmacroscopic state Claude Elwood Shannon put forward theconcept of information entropy to express the order degreeof a system [42] Let S be a set of s samples Suppose theclassification attribute has m different values Ci(i

1 2 m) and let si be the number of samples in class Cien for a given sample its total entropy isI(s1 s2 sm) minus 1113936

mi1 Pilog2(Pi) Among which Pi is the

probability that any sample belongs to CiLet an attributeA have k different values a1 a2 ak1113864 1113865

Using the attribute to divide the set S into k subsetsS1 S2 Sk1113864 1113865 Among which Sj contains the samples with avalue aj in the set S Let sij be the number of samples of classCj in subset Sjen the information entropy of the samplesdivided according to A is given by

E(A) 1113944k

j1

s1j + s2j + middot middot middot + smj

sI s1j s2j smj1113872 1113873 (2)

where I(s1j s2j smj) minus 1113936mi1 Pijlog2(Pij) And Pij

((s1j + s2j + middot middot middot + smj)s) is the probability of the samples ofclass Cj in subset Sj

Finally the information gain obtained by dividing the setS according to the attribute A is Gain(A)

I(s1 s2 sm) minus E(A) Obviously the smaller the E(A)the larger the Gain(A) which means more information isprovided by A to judge the usersrsquo purchase intention and theimportance of A is higher After normalizing the Gain(A)

values of the three DAFs the weight values β1 sim β3 in thefitness function could be formed

We sent questionnaires online to eight professional usersto investigate their purchase decisions of products formed bydifferent combinations of interaction modes e purchasedecisions were divided into two categories which were ldquobuyrdquoand ldquonot buyrdquo Delphi method was used to collect theiropinions in an all-round way and ensure the consistency oftheir decisions Compared with taking the most frequentlyselected option as the final decision the results of the Delphimethod were more scientific and more information wasavailable in the research [43]

Complexity 5

323 Genetic Operation and Evolutionary Process In thissystem the premise to debug the genetic operation andmakejudgment was due to the irrationality and uncertainty of theshape and spatial dimension design of automobile con-trollers and panels the dimension evolution driven by usersrsquoexpectation was not only to search for the optimal solutionbut also to obtain the satisfactory solution under limitedresources erefore the model was different in codingselection crossover and mutation from traditional GAwhich was to solve problems represented by explicitfunctions

(1) Coding Generally there are two kinds of codingmethods for optimization problems which are real numbercoding and binary coding And both have advantages anddisadvantages Mapping errors exist in traditional binarycoding when continuous functions are discretizedWhen theindividual coding string is short the accuracy requirementsmay not be met While when the string is long although thecoding accuracy can be improved the search space of GAwill expand dramatically [44] erefore we used realnumber coding in this study

In the process of evolution a population Pop(t) wasformed Each chromosome Xt

i in the population representeddata from one of the 30 experiments e genes in thechromosome were composed of PCD reported by thesubjects after completing the whole experiment processwhich could be expressed as follows

Pop(t) Xt1 X

t2 X

ti X

tM1113966 1113967

Xti θt

i1 θti2 θt

ib θtiN1113960 1113961

(3)

where t represented the generation number of evolutions θrepresented gene coding and M represented the populationsize

(2) Initialized Population and Selection e populationneeded to be initialized before running GA According to theresearch demand PCD data from the 30 experiments weretaken as the primary population Each time data of the 14PCD indexes were collected to form the population matrix

e selection operation was the process of selectingindividuals from the previous generation of the populationto form the next generation e purpose of selection was toobtain fine individuals based on the fitness values so thatthey would have the opportunity to reproduce as parents forthe next generation Individuals with high fitness were morelikely to be inherited while the ones with low fitness wereless likely After debugging and comparison we chose themethod of normalized geometric select (NGS) to performthe selection operation NGS was mainly to sort the fitnessvalues and the better individuals would be maintained asparents is was more suitable for the selection of productdesign schemes and also helpful to prevent better individualsfrom being damaged

(3) Crossover and Mutation By crossover operation anew generation of individuals could be obtained whichcombined the characteristics of their parents and embodiedthe idea of information exchange In the case of real number

coding the arithmetic uniform crossover (AUC) betweenindividuals was generally utilized [45] AUC was a linearcombination of two individuals to produce two new onesemethod of crossover and getting the next generation wasas follows

c1 p1lowast a + p2lowast (1 minus a)

c2 p1lowast (1 minus a) + p2lowast a(4)

where p1 and p2 were the parents and a was a random mixamount

In this study the nonuniform mutation (NM) was usedto perform the mutation operation Let an individual beX X1X2 Xk Xl If Xk was the variation point andits value range was [Uk

min Ukmax] a new individual

X X1X2 Xkprime Xl could be obtained after non-

uniformly mutating at this point And the new gene valuewas given by

Xkprime

Xk + Δ t Ukmax minus Xk1113872 1113873 if random(0 1) 0

Xk minus Δ t Xk minus Ukmin1113872 1113873 if random(0 1) 1

⎧⎨

⎩

(5)

where Δ(t y) was a random number conforming to thenonuniform distribution in the range [0 y] y representedUk

max minus Xk and Xk minus Ukmin It was required that with the

increase of evolutionary generation number t the proba-bility of Δ(t y) approaching to 0 also raised gradually

In this study Δ(t y) was defined as follows

Δ(t y) y 1 minus r(1minus tT)b

1113872 1113873 (6)

where r was a random number conforming to the non-uniform distribution in the range [0 1] which was theselection pressure T was the maximum number of evolu-tionary generations in this study T 25 In the 25 iterationsthe better individuals with high fitness could be found out toexplore the evolutionary mechanism of the appropriate HMIdimensions of the DAFs b was a parameter to adjust thevariable step size which was a system parameter It deter-mined the dependence degree of the random number dis-turbance on t and the range was usually 2sim5 Afterdebugging the model and making comparisons repeatedlythis parameter was set as b 4 in this study

4 Results

41 Information Entropy and the Weights of the DAFs

4116e Expectation Information for the Classification of theHMI of Take-Over Tasks According to different combina-tions of the seven interaction modes in Table 1 a total of2lowast 2lowast 312 HMI schemes could be generated By means ofDelphi method purchase decisions for the 12 schemes wereinvestigated After four rounds of expert investigations thepurchase decisions of the eight professional users reached anagreement e classification results are shown in Table 2

According to the purchase intentions of the professionalusers 12 HMI schemes could be divided into two categories

6 Complexity

U1 1 (buy) and U2 2 (not buy) By summarizing andanalyzing the data the probabilities of the two categorieswere as follows P (U1) (512) and P (U2) (712) Basedon the formula the total entropy was as follows

I (U) minus512log2

512

minus712log2

712

09799 (7)

412 6e Conditional Entropy and Information Gain of EachInfluencing Factor In this study there were three factorsthat determined the purchase intentions of users A1 thestarting mode of the self-adaptive cruise A2 the position ofthe function key to adjust the distance between cars and A3the position of the function key of auto hold e proba-bilities of the factors and each condition are shown inTable 3

e information entropy values of the three factors wereas follows

E A1( 1113857 612

I(1 5) +612

I(4 2) 07842

E A2( 1113857 612

I(3 3) +612

I(2 4) 09592

E A3( 1113857 412

I(1 3) +412

I(1 3) +412

I(3 1) 08113

(8)

erefore the information gains of the factors were asfollows

Gain A1( 1113857 I(U) minus E A1( 1113857 01957

Gain A2( 1113857 I(U) minus E A2( 1113857 00207

Gain A3( 1113857 I(U) minus E A3( 1113857 01686

(9)

According to the information gains the starting mode ofthe self-adaptive cruise was the most important for the userrsquosperceived comfort and purchase intension while the posi-tion of the function key to adjust the distance between carswas not so significant By normalizing the information gainsof each influencing factor the results were taken as theweights of the factors which were

β1 Gain A1( 1113857

Gain A1( 1113857 + Gain A2( 1113857 + Gain A3( 1113857 05083

β2 Gain A2( 1113857

Gain A1( 1113857 + Gain A2( 1113857 + Gain A3( 1113857 00538

β3 Gain A3( 1113857

Gain A1( 1113857 + Gain A2( 1113857 + Gain A3( 1113857 04379

(10)

After taking these values into formula (1) the artificialfitness function could be constructed as follows

F 1

05083lowastQ1 + 00538lowastQ2 + 04379lowastQ3( 1113857 (11)

42 Results of GA

421 Characteristics of the Population e population sizeof this study was Pop 30 For the convenience of recordingthe unit of each dimension was decimeter (dm)e featuresof the 14 dimensions are described in Table 4

422 Solving Process of GA To match the userrsquos expecta-tion the seed of each initial population should be able to

Table 2 Classification of purchase intentions to interaction modes

e starting mode of self-adaptive cruise

e position of the function key to adjust thedistance between cars

e position of the function key ofauto-hold

Purchasedecision

Pulling the side stalk On the middle of the left spoke Under the EPB button 2Pressing the button On the middle of the left spoke Under the EPB button 2Pulling the side stalk On the bottom edge of the left spoke Under the EPB button 2Pressing the button On the bottom edge of the left spoke Under the EPB button 1Pulling the side stalk On the middle of the left spoke On the left of the EPB button 2Pressing the button On the middle of the left spoke On the left of the EPB button 1Pulling the side stalk On the bottom edge of the left spoke On the left of the EPB button 2Pressing the button On the bottom edge of the left spoke On the left of the EPB button 2Pulling the side stalk On the middle of the left spoke On the right of the EPB button 1Pressing the button On the middle of the left spoke On the right of the EPB button 1Pulling the side stalk On the bottom edge of the left spoke On the right of the EPB button 2Pressing the button On the bottom edge of the left spoke On the right of the EPB button 1

Table 3 e probabilities of the three factors and each condition

P (Vi) P (U1 | Vi) P (U2 | Vi)

A1Side stalk 612 16 56Button 612 46 26A2Middle 612 36 36Bottom edge 612 26 46A3Under the EPB button 412 14 34On the left of the EPB button 412 14 34On the right of the EPB button 412 34 14P (Vi) represents the probability of each condition of the influencingfactors P(U1 | Vi) and P(U2 | Vi) respectively indicate the conditionalprobabilities of the two categories

Complexity 7

cover part of the information of the PCD data And it shouldbe capable of satisfying the geometric description rules ofproduct evolution in terms of space and size [46] e initialpopulation of this paper came from repeated experiments inmany days and the perceptual experience and spatial at-tributes reflected by the data were stable e optimal so-lution of the model was obtained after 25 iterations Figure 3shows the evolutionary process of GA

In the figure the red line represents the best fitness valueof each generation And the blue dotted line describes themean fitness of each generation It could be seen from thefigure that the optimal value in the model increased with thenumber of iterations and converged into a straight line afterreaching the maximum value Until the optimal value of themodel remained unchanged the global optimal solutioncould be obtained

e genetic evolution stopped after 25 iterations It couldbe seen from the figure that although the deviation betweenthe optimal fitness values and the mean ones increased afterthe 9th iteration it continued to evolve which met theexpectation of evolution After the 21st iteration the devi-ation was close to 0

In the 25 generations of evolutionary iteration based onthe PCD data eight individuals with the highest fitness wereobtained which were the HMI dimension schemes that metthe usersrsquo expectation best in a certain generation of thepopulation It would be rational to make designs of HMIssuch as steering wheels and gear lever panels in the laterstage according to these fine values is could not onlyinherit the modelling genes of the HMIs but was alsocoupled to the user expectation to a certain extent e fi-nally generated dimensional data of the individuals with thehighest fitness are shown in Table 5

43 Significance of the Difference between the Original Di-mensions and the Evolved Ones To further validate whetherthe individuals with the highest fitness could bring about a

better take-over performance it was reasonable to collect theneeded time to complete driving assistant tasks under theHMI with the original sizes and the one with the evolvedsizes And the differences between the two sets of HMIdimensions could be analyzed

In the last of the 30 experiments we collected the time ofthe 13 subjects to complete the take-over tasks under theseven interaction modes in Table 1 Since the tasks had beenrepeated many times the subjects were skilled and the taskcompletion time would be relatively stable After making 3Dprinting models of the side stalk and the buttons accordingto the most suitable dimensions (Table 5) the drivingsimulator was reformeden the 13 subjects were invited torepeat the seven tasks with the new HMI and this set ofcompletion time was collected To judge whether the di-mensions could effectively improve the interaction perfor-mance and efficiency of driving information processingpaired-sample t-test was used on the two groups of data eresults are shown in Table 6

It could be seen from the results that except for A2 andC2 the completion time under the two sets of dimensions inthe process of the other five tasks all had a significant dif-ference (Figure 4)

Figure 4 shows that the completion time of each task wasshorter when the dimensions with high fitness were used onthe HMI is indicated that the driverrsquos information pro-cessing was faster which helped to improve the quality oftake-over and the ability to control the car As the driverrsquosattention to the HMI might compete with the execution ofdriving tasks for cognitive resources [47] and impair thedriving performance a shorter completion time of take-overtasks was conducive to the distribution of the driverrsquospsychological resources and more beneficial to trafficefficiency

It could be found by comparing the completion time ofeach task under the new HMI that (A) for starting the self-

Table 4 Features of the 14 dimensions

Tasks anddimensions

Mean(dm)

Stddeviation

Minimum(dm)

Maximum(dm)

A1 a11 19513 00789 18086 2091a12 06055 0063 05117 06977

A2 a21 08624 00242 08224 08979a22 00784 00477 00008 01487

B1 b11 13356 00633 12528 14475b12 00315 00169 00024 00588

B2 b21 13999 00526 13060 14952b22 0322 00139 03011 03497

C1 c11 29776 01254 28032 31940c12 3272 01414 30207 34928

C2 c21 27451 00305 27010 27966c22 32714 01304 30143 34819

C3 c31 32929 00596 32019 33963c32 3244 01523 30011 34932

2500735

0074

00745

0075

00755

0076

00765

0077

00775

0078

Fittn

ess

5 10 15 200Generation

Mean fitnessBest fitness

Figure 3 Solving process of GA

8 Complexity

adaptive cruise the completion time of operating with a sidestalk was shorter (B) for the task of adjusting the distancebetween cars the completion time would be shorter whenoperating with a button located in the middle of the steeringwheel spoke (C) for the task of auto hold a starting buttonon the left side of the EPB button could bring about a shortercompletion time ese results provided a reference for thefuture automobile interior design

5 Discussion

Driving assistant system (DAS) plays an important role inthe context of human-computer codriving As reportedDAS can help the driver to complete the intense drivingtasks effectively and abate the pressure level of the driver[48] erefore it is meaningful to make a study for thecontroller dimensions in the DAS e study collectedcomfortable dimensions perceived by the drivers in a relaxedstate e aim was to detect a set of reasonable dimensions

which could make drivers at a low level of stress In ad-dition it is valuable for drivers to adapt to the different roadenvironment and realize sustainable traffic to optimize thesizes of interactive components on the in-vehicle HMI emain goals of in-vehicle technologies and cooperativeservices were to reduce traffic congestion and improve roadsafety [4]

e study attempted to analyze the optimal HMI di-mensions suitable for take-over tasks e involved HMIcomponents such as the steering wheel self-adaptive cruiselever distance adjustment button and auto hold buttonwere all direct design elements for users to see and touchBecause the design and performance of a product wereimportant factors affecting the decision-making of con-sumers [49] we calculated the information entropy andinformation gains of the three kinds of take-over tasks forclassifying usersrsquo purchase intentions e informationtransmitted by a product was multivariant complex andfuzzy so that there would be dissipation in the process oftransmission erefore the image evaluation process ofproducts could be regarded as a chaotic state e feedbackinformation could be calculated by information entropy tocarry out a comprehensive evaluation e smaller the en-tropy of an image the clearer it was From the perspective ofchaos degree the 2nd factor in Table 3 (the position of thefunction key to adjust the distance between cars) broughtmore uncertainty to the purchase decision(E(A2) 09592) which indicated that little informationwas provided by the 2nd factor for judging the usersrsquo de-cisions On the contrary the uncertainty produced by the 1stfactor (the starting mode of the self-adaptive cruise) was thesmallest (E(A1) 07842) is illustrated that using theside stalk or a button to complete the switch of the controlright had a great impact on the satisfaction and purchasedecision of users

After normalizing the information entropy of threekinds of take-over tasks their weight values were calculated

Table 5 e finally generated dimensional data of the individuals with the highest fitness

(A) e starting mode of self-adaptive cruise

(B) e position of thefunction key to adjust thedistance between cars

(C) e position of the function key of auto-hold Fitness value

a11 a12 a21 a22 b11 b12 b21 b22 c11 c12 c21 c22 c31 c32202 062 089 009 136 004 148 030 284 302 271 338 321 313 008

Table 6 e paired sample t-test results of task completion time under the two HMIs with different dimensions

Interaction modes of take-over tasks

e task completiontime under the original

dimensions (s)

e task completiontime under the evolved

dimensions (s) t-value Sig

Mean Std dev Mean Std devA1lowast 0715 0185 0537 0167 2553 0025A2 0657 0181 0637 0186 1151 0272B1lowast 0648 013 0462 0206 2797 0016B2lowastlowast 0736 0106 0571 0124 3116 0009C1lowast 0753 0295 0562 0222 2515 0027C2 0667 0226 0522 0176 1999 0069C3lowast 069 0208 0537 0225 2355 0036

A1lowast B1lowast B2lowastlowast C1lowast C3lowast

0

02

04

06

08

1

12

Time under the original dimensionsTime under the evolved dimensions

Figure 4 e completion time of the tasks with significant dif-ferences under the two HMIs with different dimensions

Complexity 9

and the fitness function was constructed In this way thedimensions with the highest fitness could be found out bymeans of GA

e background and driving experience of the subjects weredifferent and their familiarity with the take-over tasks and theHMI was variant Besides the noise of usersrsquo subjective per-ception data was always high Taking the above factors intoconsideration we collected PCD data repeatedly from the same13 subjects to form the initial population From the evolutionarytrajectory it could be found that in the 8th 10th 11th 14th 17thand 21st iteration there appeared an optimal solution

GA was always used for parameter optimization bysearching the optimal solutions while seldom used in userresearch and environment perception analysis But in studiesof user-centred design it could search the set of satisfactorysolutions driven by user expectations [46] Many pieces ofresearch had utilized GA to improve the ergonomic effi-ciency or optimize interactive component sizes in an op-erating environment [34ndash38 50 51] In this study wereported a driver perception research which investigated themost comfortable HMI sizes under different interactionmodes of take-over tasks By means of normalized geometricselection arithmetic uniform crossover and nonuniformmutation GA was applied in the research of perceivedcomfortable dimensions and a set of reasonable dimensionswas obtained e dimensional evolutionary method dis-cussed in this study embodied the concept of dynamic in-teraction design which took the mutual influence andrestrictions among different dimensions into considerationCompared with traditional methods of measuring manytimes and taking the average this method was more ad-vanced And the study enriched theoretical researches in thefield of product design and collaborative optimization

Perceived comfort was a complex problem for its dif-ficulty to be quantified objectively But it was useful in er-gonomic explorations Researchers had found that theperceptual messages from product property and usage re-quirements were competent to the analysis of user experi-ence [52] erefore it was reasonable to probe suitabledimensions by perceived comfort For the study of HMIdesign effect in a specific environment these data couldproduce important and dependable results [53] Rationaldimensions of the automobile interior HMI played an im-portant role in the driverrsquos operation quality Caruso [54]studied driversrsquo experience on the seat by an experiment andfound that interior space influenced the driversrsquo ergonomicperformance [54] For the take-over tasks in the situation ofhuman-computer codriving the dimensions also affectedwhether the driver could safely achieve the control rightswitch in the shortest possible time

In this paper paired-sample t-test was used to testwhether the sizes with the highest fitness could bring about asignificant reduction in task completion time In the resultswe found that when the interface adopted the finally evolvedsizes the completion time of each task was shorter As thecompletion time of in-vehicle interactive tasks was related tothe vehicle speed and driver errors under the context ofhuman-computer codriving [21 55] a shorter completiontime would help the driver to control the speed better

6 Limitation

e study explored the reasonable dimensions of the hu-man-machine interface in intelligent cars Neverthelessthere were still some deficiencies in the equipment used inthis study e driving simulator was easier to control andreform and helped to reduce the subjectsrsquo pressure How-ever the immersion visual interactivity and fidelity degreewere not high enough After obtaining reliable conclusionsin this study outdoor experiment with a real vehicle could beconducted in the next stage

Besides the sample size of this study was limited Whenthe sample size was larger in future studies the represen-tativeness of the data would be better and the results of GAcould be further improved

Finally usersrsquo perception of the HMI dimensions de-viated from each other In the next phase of the studyelectroencephalography (EEG) signal in the process ofperforming the driving tasks could be used By this meansthe subjectsrsquo attention and intention when interacting withthe HMI could be reflected in an objective way [56 57] as asupplement to their subjective report

7 Conclusion

For HMI design of industrial products physical concretemodelling parameters such as the product shape size andlayout were the objective properties that users could directlyperceive and the basis for the product to be expressed Basedon the dynamic measurement principle this paper proposedan optimized design method for ergonomic dimensions byGA e results indicated that the method was effective andprovided a new way for product shape innovation andhuman-machine dimensional automatic design e mainconclusions of this study were as follows

(1) Based on the information entropy theory the weightvalues of the three driving assistance functions werecalculated e results showed that the informationgain of the starting mode of the self-adaptive cruisewas highest while the information gain of the po-sition of the function key to adjust the distancebetween cars was lowest which provided a theo-retical basis for the dimensional fitness evaluationmechanism of the HMI

(2) e evolutionary process of HMI dimensionsdriven by user expectation was studied and theevolution conditions were analyzed e studyconfirmed the feasibility of normalized geometricselection arithmetic uniform crossover and non-uniform mutation in iterative researches for hu-man-machine dimensions rough parameterdebugging this evolutionary method was proved tobe effective in obtaining appropriate sizes of HMIcomponents e model could make full use of theadvantages of GA to improve the user researchperformance such as few restrictions on the opti-mization high robustness and an easily achievedglobal search

10 Complexity

(3) After finding the dimensions with the highest fitnessand applying them to the design of the HMI within-subjects experiments showed that the completiontime of five take-over tasks was significantly shorterexcept for that of Task A2 and C2is indicated thatthe information processing of drivers under the HMIwith evolved dimensions was faster which was morebeneficial to the vehicle control and take-overquality Intelligent vehicles achieved intelligent in-formation exchange with people vehicles and roadsthrough in-vehicle sensor system and informationterminals erefore the evolved dimensions couldmake the vehicle apt to respond to the environmentand have more time to fully deal with the complextraffic environment and would help to improve thetraffic efficiency therefrom

Data Availability

e data used to support the findings of this study areavailable from the corresponding author upon request

Consent

e study respected and guaranteed the subjectsrsquo right toparticipate in the study and allowed them to withdraw from thestudy unconditionally at any stage e study strictly followedthe informed consent procedure e subjects were asked tosign a consent form to inform them of the purpose processduration of the experiment matters needing attention and themethods and purpose of collecting the data so as to reducetheir anxiety and confusion In the experiment the personalsafety and health of the subjects took priority According to theusage guidelines of the driving simulator we tried to avoid anypossible injury in driving tasks e privacy of the subjects wasstrictly protected e subjects were informed of the storageuse and confidentiality measures of their personal informationand image data in advance truthfully And we would notdisclose the information to a third party or spread them on theInternet without authorization

Conflicts of Interest

e authors declare that they have no conflicts of interest

Acknowledgments

is researchwas funded by BeijingUrbanGovernance ResearchProject (grant no 20XN244) Scientific Research Program ofBeijing Education Commission (grant no KM202010009003)Yuyou Talent Support Program of North China University ofTechnology (grant no 107051360018XN012018) Tianjin YouthTalents Reserve Support Program and Chinese ErgonomicsSociety amp Kingfar Joint Research Fund for Outstanding YoungScholars (grant number CES-Kingfar-2019-001)

References

[1] S Jin J Yang E Wang and J Liu ldquoe influence of high-speed rail on ice-snow tourism in north eastern ChinardquoTourism Management vol 78 Article ID 104070 2020

[2] J Yang R Yang J Sun T Huang and Q Ge ldquoe spatialdifferentiation of the suitability of ice-snow tourist destina-tions based on a comprehensive evaluation model in ChinardquoSustainability vol 9 no 5 Article ID 774 2017

[3] W-Y Chung T-W Chong and B-G Lee ldquoMethods to detectand reduce driver stress a reviewrdquo International Journal ofAutomotive Technology vol 20 no 5 pp 1051ndash1063 2019

[4] H Farah and H N Koutsopoulos ldquoDo cooperative systemsmake driversrsquo car-following behavior saferrdquo TransportationResearch Part C Emerging Technologies vol 41 pp 61ndash722014

[5] SAE Taxonomy and Definitions for Terms Related to DrivingAutomation Systems for On-Road Motor Vehicles SAEJ3016TM Warrendale PA USA 2018

[6] J Hoegg J W Alba and D W Dahl ldquoe good the bad andthe ugly influence of aesthetics on product feature judg-mentsrdquo Journal of Consumer Psychology vol 20 no 4pp 419ndash430 2010

[7] M Kyriakidis R Happee and J C F De Winter ldquoPublicopinion on automated driving results of an internationalquestionnaire among 5000 respondentsrdquo TransportationResearch Part F Traffic Psychology and Behaviour vol 32pp 127ndash140 2015

[8] F Meng and C Spence ldquoTactile warning signals for in-vehiclesystemsrdquo Accident Analysis amp Prevention vol 75 pp 333ndash346 2015

[9] J Wan and C Wu ldquoe effects of lead time of take-overrequest and nondriving tasks on taking-over control of au-tomated vehiclesrdquo IEEE Transactions on Human MachineSystems vol 48 no 6 pp 582ndash591 2018

[10] S Petermeijer P Hornberger I Ganotis J D Winter andK Bengler ldquoe design of a vibrotactile seat for conveyingtake-over requests in automated drivingrdquo in Proceedings of the8th International Conference on Applied Human Factors andErgonomics Los Angeles CA USA July 2017

[11] T Ito A Takata and K Oosawa ldquoTime required for take-overfrom automated to manual drivingrdquo in Proceedings of the SAE2016 World Congress and Exhibition Detroit MI USA April2016

[12] K Zeeb A Buchner and M Schrauf ldquoWhat determines thetake-over time An integrated model approach of driver take-over after automated drivingrdquo Accident Analysis amp Preven-tion vol 78 pp 212ndash221 2015

[13] B Mok M Johns K J Lee et al ldquoEmergency automation offunstructured transition timing for distracted drivers of au-tomated vehiclesrdquo in Proceedings of the IEEE InternationalConference on Intelligent Transportation Systems Las PalmasSpain September 2015

[14] S Samuel A Borowsky S Zilberstein and D L FisherldquoMinimum time to situation awareness in scenarios involvingtransfer of control from an automated driving suiterdquoTransportation Research Record Journal of the TransportationResearch Board vol 2602 no 2602 pp 115ndash120 2016

[15] W Vlakveld N van Nes J de Bruin L Vissers andM van der Kroft ldquoSituation awareness increases when drivershave more time to take over the wheel in a level 3 automated

Complexity 11

car a simulator studyrdquo Transportation Research Part F TrafficPsychology and Behaviour vol 58 pp 917ndash929 2018

[16] N Merat A H Jamson F C H Lai M Daly andO M J Carsten ldquoTransition to manual driver behaviourwhen resuming control from a highly automated vehiclerdquoTransportation Research Part F Traffic Psychology and Be-haviour vol 27 no 26 pp 274ndash282 2014

[17] C Z Wu H R Wu and N C Lyu ldquoReview of control switchand safety of human-computer driving intelligent vehiclerdquoJournal of Traffic and Transportation Engineering vol 18no 6 pp 131ndash141 2018

[18] C Gold D Dambock L Lorenz and K Bengler ldquoldquoTakeoverrdquo how long does it take to get the driver back into thelooprdquo Proceedings of the Human Factors and ErgonomicsSociety Annual Meeting vol 57 no 1 pp 1938ndash1942 2013

[19] H Yang Y Zhao and Y Wang ldquoIdentifying modeling formsof instrument panel system in intelligent shared cars a studyfor perceptual preference and in-vehicle behaviorsrdquo Envi-ronmental Science and Pollution Research vol 27 no 1pp 1009ndash1023 2020

[20] A Berkovich A Lu B Levine and A V Reddy ldquoObservedcustomer seating and standing behavior and seat preferenceson board subway cars in New York cityrdquo TransportationResearch Record Journal of the Transportation ResearchBoard vol 2353 no 1 pp 33ndash46 2013

[21] R Li Y V Chen C Sha and Z Lu ldquoEffects of interface layouton the usability of in-vehicle information systems and drivingsafetyrdquo Displays vol 49 pp 124ndash132 2017

[22] H Kim S Kwon J Heo H Lee and M K Chung ldquoe effectof touch-key size on the usability of in-vehicle informationsystems and driving safety during simulated drivingrdquo AppliedErgonomics vol 45 no 3 pp 379ndash388 2014

[23] H Kim and H Song ldquoEvaluation of the safety and usability oftouch gestures in operating in-vehicle information systemswith visual occlusionrdquo Applied Ergonomics vol 45 no 3pp 789ndash798 2014

[24] M Koerber C Gold D Lechner and K Bengler ldquoe in-fluence of age on the take-over of vehicle control in highlyautomated drivingrdquo Transportation Research Part F TrafficPsychology amp Behaviour vol 39 pp 19ndash32 2016

[25] H Clark A C Mclaughlin B Williams and J Feng ldquoPer-formance in takeover and characteristics of non-driving re-lated tasks during highly automated driving in younger andolder driversrdquo Proceedings of the Human Factors amp Ergo-nomics Society Annual Meeting vol 61 no 1 pp 37ndash41 2017

[26] I Lijarcio S A Useche J Llamazares and L MontoroldquoPerceived benefits and constraints in vehicle automationdata to assess the relationship between driverrsquos features andtheir attitudes towards autonomous vehiclesrdquo Data in Briefvol 27 Article ID 104662 2019

[27] K-Z Chen and X-A Feng ldquoVirtual genes of manufacturingproducts and their reforms for product innovative designrdquoProceedings of the Institution of Mechanical Engineers Part CJournal of Mechanical Engineering Science vol 218 no 5pp 557ndash574 2004

[28] K-Z Chen X-A Feng and X-C Chen ldquoReverse deductionof virtual chromosomes of manufactured products for theirgene-engineering-based innovative designrdquo Computer-AidedDesign vol 37 no 11 pp 1191ndash1203 2005

[29] H-C Tsai and J-R Chou ldquoAutomatic design support andimage evaluation of two-coloured products using colour as-sociation and colour harmony scales and genetic algorithmrdquoComputer-Aided Design vol 39 no 9 pp 818ndash828 2007

[30] S-W Hsiao C-F Hsu and K-W Tang ldquoA consultation andsimulation system for product color planning based on in-teractive genetic algorithmsrdquo Color Research amp Applicationvol 38 no 5 pp 375ndash390 2013

[31] Y Sun W-l Wang Y-w Zhao and X-j Liu ldquoUser imageoriented interactive genetic algorithm evaluation mode inproduct developmentrdquo Computer Integrated ManufacturingSystems vol 18 no 2 pp 276ndash281 2012

[32] R Kamalian E Yeh Z Ying A M Agogino and H TakagildquoReducing human fatigue in interactive evolutionary com-putation through fuzzy systems and machine learning sys-temsrdquo in Proceedings of the 2006 IEEE InternationalConference on Fuzzy Systems Vancouver Canada July 2006

[33] Z Gu M Xi Tang and J H Frazer ldquoCapturing aestheticintention during interactive evolutionrdquo Computer-AidedDesign vol 38 no 3 pp 224ndash237 2006

[34] G Harih M Borovinsek and Z Ren Optimisation ofProductrsquos Hand-Handle Interface Material Parameters forImproved Ergonomics Springer International PublishingCham Switzerland 2015

[35] P Cheng D Chen and J Wang ldquoClustering of the bodyshape of the adult male by using principal component analysisand genetic algorithm-BP neural networkrdquo Soft Computingvol 24 no 17 Article ID 13219 2020

[36] N M B Serrano P J G Nieto A S Sanchez F S Lasherasand P R Fernandez AHybrid Algorithm for the Assessment ofthe Influence of Risk Factors in the Development of Upper LimbMusculoskeletal Disorders Springer Cham Switzerland 2018

[37] S S Sankar O-M Holman A F Gazabon and C J AcevedoldquoApplication of genetic algorithm to job scheduling underergonomic constraints in manufacturing industryrdquo Journal ofAmbient Intelligence and Humanized Computing vol 10no 5 pp 2063ndash2090 2019

[38] A-Z Atiya L Lee and K Xing ldquoDeveloping a multi-ob-jective genetic optimisation approach for an operationaldesign of a manual mixed-model assembly line with walkingworkersrdquo Journal of Intelligent Manufacturing vol 27 no 5pp 1ndash17 2014

[39] P Jedlicka and T Ryba ldquoGenetic algorithm application inimage segmentationrdquo Pattern Recognition and Image Anal-ysis vol 26 no 3 pp 497ndash501 2016

[40] L V Campenhout J Frens K Overbeeke A Standaert andH Peremans ldquoPhysical interaction in a dematerializedworldrdquo International Journal of Design vol 7 no 1 pp 1ndash182013

[41] K Nandhini and S R Balasundaram ldquoImproving readabilitythrough individualized summary extraction using interactivegenetic algorithmrdquo Applied Artificial Intelligence vol 30no 7 pp 635ndash661 2016

[42] R M Gray Entropy and Information 6eory Springer NewYork NY USA 2nd edition 2011

[43] B Anderhofstadt and S Spinler ldquoFactors affecting the pur-chasing decision and operation of alternative fuel-poweredheavy-duty trucks in Germanymdasha Delphi studyrdquo Trans-portation Research Part D Transport and Environmentvol 73 pp 87ndash107 2019

[44] X-m Wu and Y Feng ldquoReactive power optimization ofdistribution network based on improved genetic algorithmrdquoJournal of Xirsquoan Shiyou University (Natural Science Edition)vol 30 no 3 pp 95ndash99 2015

[45] Z Ling andM-j Li ldquoState space evolutionary algorithm basedon non-uniform mutation operatorrdquo Computer Technologyand Development vol 28 no 9 pp 68ndash71 2018

12 Complexity

[46] W Hu and J Zhao ldquoAutomobile styling gene evolutiondriven by usersrsquo expectation imagerdquo Journal of MechanicalEngineering vol 47 no 16 pp 176ndash181 2011

[47] J J Scott and R Gray ldquoA comparison of tactile visual andauditory warnings for rear-end collision prevention in sim-ulated drivingrdquo Human Factors 6e Journal of the HumanFactors and Ergonomics Society vol 50 no 2 pp 264ndash2752008

[48] B Reimer B Mehler and J F Coughlin ldquoReductions in self-reported stress and anticipatory heart rate with the use of asemi-automated parallel parking systemrdquo Applied Ergonom-ics vol 52 pp 120ndash127 2016

[49] B DrsquoIppolito ldquoe importance of design for firms com-petitiveness a review of the literaturerdquo Technovation vol 34no 11 pp 716ndash730 2014

[50] S Asensio-Cuesta J A Diego-Mas L Canos-Daros andC Andres-Romano ldquoA genetic algorithm for the design of jobrotation schedules considering ergonomic and competencecriteriardquo International Journal of Advanced ManufacturingTechnology vol 60 no 9ndash12 pp 1161ndash1174 2012

[51] S-C Lee H-E Tseng C-C Chang and Y-M HuangldquoApplying interactive genetic algorithms to disassembly se-quence planningrdquo International Journal of Precision Engi-neering and Manufacturing vol 21 no 4 pp 663ndash679 2020

[52] M G Shishaev V V Dikovitsky and L V Lapochkina ldquoeexperience of building cognitive user interfaces of multido-main information systems based on the mental model ofusersrdquo in Proceedings of the 6th Computer Science On-LineConference (CSOC 2017) Prague Czech Republic April 2017

[53] S Lee H Alzoubi and S Kim ldquoe effect of interior designelements and lighting layouts on prospective occupantsrsquoperceptions of amenity and efficiency in living roomsrdquo Sus-tainability vol 9 no 7 Article ID 1119 2017

[54] G Caruso ldquoMixed reality system for ergonomic assessment ofdriverrsquos seatrdquo International Journal of Virtual Reality vol 10no 2 pp 69ndash79 2011

[55] M L Reyes and J D Lee ldquoEffects of cognitive load presenceand duration on driver eye movements and event detectionperformancerdquo Transportation Research Part F Traffic Psy-chology and Behaviour vol 11 no 6 pp 391ndash402 2008

[56] D Cisler P M Greenwood D M Roberts R McKendrickand C L Baldwin ldquoComparing the relative strengths of EEGand low-cost physiological devices in modeling attentionallocation in semiautonomous vehiclesrdquo Frontiers in HumanNeuroscience vol 13 Article ID 109 2019

[57] I-H Kim J-W Kim S Haufe and S-W Lee ldquoDetection ofbraking intention in diverse situations during simulateddriving based on EEG feature combinationrdquo Journal of NeuralEngineering vol 12 no 1 Article ID 016001 2014

Complexity 13

(2) After stopping stably the subjects ought to startmanual driving again

(3) When the system first prompted that the self-adaptive cruise function could be used the subjectsturned the cruise control stalk D is was to sim-ulate switching the manual driving mode to theautomated driving mode and entering the state ofself-driving At this moment the virtual vehicle inthe system would run independently according tothe preset navigation route

(4) When the system prompted for adjusting the dis-tance between cars for the first time the subjectsneeded to press the M button

(5) en the subjects should complete auto hold self-adaptive cruise and distance adjustment in sequence

for the second time and pressed the buttons B E andS respectively

(6) Complete the auto hold operation for the third timeand press the C button

(7) After the completion of all the above tasks thesubjects were asked to point out the comfortableposition of the side stalk and buttons according tothe operating experience A researcher recorded thepoint with a marker pen on the simulator andcalibrated the sizese experiment used a within-subjects design erewere 13 subjects Each experiment lasted about 3minutes and the subjects needed to carry out theabove tasks five times to complete the experiment ofone day In order to ensure the accuracy and stability

Table 1 e driving assistance functions interaction modes and the analyzed dimensions

Driving assistancefunctions Interaction modes of take-over tasks Dimensions between the interactive component and the

steering wheel center

(A) Self-adaptive cruise

(A1) Switch the control right by the side stalk D

(a11) e PCD between the left endpoint of the stalk Dand the vertical centerline of the steering wheel (VCSW)(a12) e PCD between the left endpoint of the stalk Dand the horizontal centerline of the steering wheel(HCSW)

(A2) Switch the control right by the button E

(a21) e PCD between the vertical centerline (VC) ofthe button E and the VCSW(a22) e PCD between the horizontal centerline (HC)of the button E and the HCSW

(B) Adjustment of thedistance between cars

(B1) Adjust the distance by the button M which wasin the middle of the left spoke of the steering wheel

(b11)e PCD between the VC of the button M and theVCSW(b12)e PCD between the HC of the buttonM and theHCSW

(B2) Adjust the distance by the button S which wason the bottom edge of the left spoke of the steeringwheel

(b21) e PCD between the VC of the button S and theVCSW(b22) e PCD between the HC of the button S and theHCSW

(C) Auto hold

(C1) Start the function by button A which was underthe electric parking brake (EPB) button

(c11) e PCD between the VC of button A and theVCSW(c12) e PCD between the HC of button A and theHCSW

(C2) Start the function by button B which was on theleft of the EPB button

(c21) e PCD between the VC of button B and theVCSW(c22) e PCD between the HC of button B and theHCSW

(C3) Start the function by button C which was on theright of the EPB button

(c31) e PCD between the VC of button C and theVCSW(c32) e PCD between the HC of button C and theHCSW

0 1 2 3 4 5 6 7Time

Press thebutton C

Press thebutton S

Press thebutton E

Press thebutton B

Press thebutton M

Press thebutton A

Pull the sidestalk D

Start thesimulator in

manual drivingmode

None 1st auto hold 1st self-adaptivecruise

2nd self-adaptivecruise

1st adjustmentof the distancebetween cars

2nd adjustmentof the distancebetween cars

2nd auto hold 3rd auto hold

Figure 2 e task flow of the experiments

4 Complexity

of the perceptual experience the experiment wouldbe in progress for six days As a result each subjectwould carry out the tasks for 5lowast 6 30 times ePCD data collected each time were taken as the initialpopulation and the dimensions with the highestfitness would be obtained by GA

(8) In the 30th experiment the completion time anderror times of the subjects during the seven take-overtasks were collected

(9) After obtaining the PCD data with the highestadaptability models of the side stalk and buttonswere made by 3D printers to replace the originalones e subjects were asked to repeat the seventasks in Figure 2 again with the modified simulatorAnd their completion time and error times under thenew HMI would be collected

32 Information Entropy and Genetic Algorithm GA is aself-adaptive global optimization algorithm which imitatesthe natural selection and individual heredity of the biologicalworld [39] Traditionally interactive genetic algorithm(IGA) was widely used in product modelling design forsubject recognition [40] However IGA has some short-comings including the uncertainty of individual fitnessvalues the nonpersistence of the individual evaluationprocess and the nonuniqueness of optimization results [41]In this study GA was taken as the main research methodwhile PSC was used in the construction of fitness functionwhich integrated user preference intuition emotion andpsychological optimization into it erefore the fitnessvalues of evolutionary individuals could make the pop-ulation evolve in the direction that users expected

321 Fitness Function Whether the population can evolvein the direction of the ideal solution depends on the es-tablishment of fitness function to a great extent Accordingto the userrsquos PSC and the weight of the three DAFs thatneeded the driver to take over the artificial fitness functionwas formulated as follows

F 1

β1Q1 + β2Q2 + β3Q3( 1113857 (1)

Among which β1 sim β3 represented the importanceweights of the three DAFs which were obtained by calcu-lating the information entropy Q1 sim Q3 meant the com-fortable dimension of the operating components of eachDAF listed in Table 1 Q1 a11 lowast a12 + a21 lowast a22 Q2 b11 lowastb12 + b21 lowast b22 Q3 c11 lowast c12 + c21 lowast c22+ c31 lowast c32 emeanings of a11 sim c32 values are shown in Table 1 whichwere obtained in step (7) of the experiment For each in-teraction mode the comfortable position of the corre-sponding operation component was identified with a pointat first and then the distances between the point and thehorizontal and vertical centerlines of the steering wheel weremeasurede two distances were multiplied to get a relativecomfort area F was used to evaluate the fitness degree of

individuals With a smaller size the userrsquos behavioral path tocomplete the operations could be shorter and the taskcompletion time would be relatively less which were con-ducive to the completion of the take-over tasks ereforethe larger the F value was the more reasonable the sizeswere

322 Information Entropy and the Weight of the DrivingAssistance Functions Based on the entropy theory the studyexplored the importance level of the main DAFs in L1-L2automated driving and analyzed the influencing factors ofusersrsquo purchase intentions e uncertainty degree of thedivided sample set which was measured by means of cal-culating the information gain was used as the standard toweigh the quality of the division e larger the informationgain the less the uncertainty degree of the sample set

Entropy refers to the degree of system chaos which is ameasurement of the possibility of the system in a certainmacroscopic state Claude Elwood Shannon put forward theconcept of information entropy to express the order degreeof a system [42] Let S be a set of s samples Suppose theclassification attribute has m different values Ci(i

1 2 m) and let si be the number of samples in class Cien for a given sample its total entropy isI(s1 s2 sm) minus 1113936

mi1 Pilog2(Pi) Among which Pi is the

probability that any sample belongs to CiLet an attributeA have k different values a1 a2 ak1113864 1113865

Using the attribute to divide the set S into k subsetsS1 S2 Sk1113864 1113865 Among which Sj contains the samples with avalue aj in the set S Let sij be the number of samples of classCj in subset Sjen the information entropy of the samplesdivided according to A is given by

E(A) 1113944k

j1

s1j + s2j + middot middot middot + smj

sI s1j s2j smj1113872 1113873 (2)

where I(s1j s2j smj) minus 1113936mi1 Pijlog2(Pij) And Pij

((s1j + s2j + middot middot middot + smj)s) is the probability of the samples ofclass Cj in subset Sj

Finally the information gain obtained by dividing the setS according to the attribute A is Gain(A)

I(s1 s2 sm) minus E(A) Obviously the smaller the E(A)the larger the Gain(A) which means more information isprovided by A to judge the usersrsquo purchase intention and theimportance of A is higher After normalizing the Gain(A)

values of the three DAFs the weight values β1 sim β3 in thefitness function could be formed

We sent questionnaires online to eight professional usersto investigate their purchase decisions of products formed bydifferent combinations of interaction modes e purchasedecisions were divided into two categories which were ldquobuyrdquoand ldquonot buyrdquo Delphi method was used to collect theiropinions in an all-round way and ensure the consistency oftheir decisions Compared with taking the most frequentlyselected option as the final decision the results of the Delphimethod were more scientific and more information wasavailable in the research [43]

Complexity 5

323 Genetic Operation and Evolutionary Process In thissystem the premise to debug the genetic operation andmakejudgment was due to the irrationality and uncertainty of theshape and spatial dimension design of automobile con-trollers and panels the dimension evolution driven by usersrsquoexpectation was not only to search for the optimal solutionbut also to obtain the satisfactory solution under limitedresources erefore the model was different in codingselection crossover and mutation from traditional GAwhich was to solve problems represented by explicitfunctions

(1) Coding Generally there are two kinds of codingmethods for optimization problems which are real numbercoding and binary coding And both have advantages anddisadvantages Mapping errors exist in traditional binarycoding when continuous functions are discretizedWhen theindividual coding string is short the accuracy requirementsmay not be met While when the string is long although thecoding accuracy can be improved the search space of GAwill expand dramatically [44] erefore we used realnumber coding in this study

In the process of evolution a population Pop(t) wasformed Each chromosome Xt

i in the population representeddata from one of the 30 experiments e genes in thechromosome were composed of PCD reported by thesubjects after completing the whole experiment processwhich could be expressed as follows

Pop(t) Xt1 X

t2 X

ti X

tM1113966 1113967

Xti θt

i1 θti2 θt

ib θtiN1113960 1113961

(3)

where t represented the generation number of evolutions θrepresented gene coding and M represented the populationsize

(2) Initialized Population and Selection e populationneeded to be initialized before running GA According to theresearch demand PCD data from the 30 experiments weretaken as the primary population Each time data of the 14PCD indexes were collected to form the population matrix

e selection operation was the process of selectingindividuals from the previous generation of the populationto form the next generation e purpose of selection was toobtain fine individuals based on the fitness values so thatthey would have the opportunity to reproduce as parents forthe next generation Individuals with high fitness were morelikely to be inherited while the ones with low fitness wereless likely After debugging and comparison we chose themethod of normalized geometric select (NGS) to performthe selection operation NGS was mainly to sort the fitnessvalues and the better individuals would be maintained asparents is was more suitable for the selection of productdesign schemes and also helpful to prevent better individualsfrom being damaged

(3) Crossover and Mutation By crossover operation anew generation of individuals could be obtained whichcombined the characteristics of their parents and embodiedthe idea of information exchange In the case of real number

coding the arithmetic uniform crossover (AUC) betweenindividuals was generally utilized [45] AUC was a linearcombination of two individuals to produce two new onesemethod of crossover and getting the next generation wasas follows

c1 p1lowast a + p2lowast (1 minus a)

c2 p1lowast (1 minus a) + p2lowast a(4)

where p1 and p2 were the parents and a was a random mixamount

In this study the nonuniform mutation (NM) was usedto perform the mutation operation Let an individual beX X1X2 Xk Xl If Xk was the variation point andits value range was [Uk

min Ukmax] a new individual

X X1X2 Xkprime Xl could be obtained after non-

uniformly mutating at this point And the new gene valuewas given by

Xkprime

Xk + Δ t Ukmax minus Xk1113872 1113873 if random(0 1) 0

Xk minus Δ t Xk minus Ukmin1113872 1113873 if random(0 1) 1

⎧⎨

⎩

(5)

where Δ(t y) was a random number conforming to thenonuniform distribution in the range [0 y] y representedUk

max minus Xk and Xk minus Ukmin It was required that with the

increase of evolutionary generation number t the proba-bility of Δ(t y) approaching to 0 also raised gradually

In this study Δ(t y) was defined as follows

Δ(t y) y 1 minus r(1minus tT)b

1113872 1113873 (6)

where r was a random number conforming to the non-uniform distribution in the range [0 1] which was theselection pressure T was the maximum number of evolu-tionary generations in this study T 25 In the 25 iterationsthe better individuals with high fitness could be found out toexplore the evolutionary mechanism of the appropriate HMIdimensions of the DAFs b was a parameter to adjust thevariable step size which was a system parameter It deter-mined the dependence degree of the random number dis-turbance on t and the range was usually 2sim5 Afterdebugging the model and making comparisons repeatedlythis parameter was set as b 4 in this study

4 Results

41 Information Entropy and the Weights of the DAFs

4116e Expectation Information for the Classification of theHMI of Take-Over Tasks According to different combina-tions of the seven interaction modes in Table 1 a total of2lowast 2lowast 312 HMI schemes could be generated By means ofDelphi method purchase decisions for the 12 schemes wereinvestigated After four rounds of expert investigations thepurchase decisions of the eight professional users reached anagreement e classification results are shown in Table 2

According to the purchase intentions of the professionalusers 12 HMI schemes could be divided into two categories

6 Complexity

U1 1 (buy) and U2 2 (not buy) By summarizing andanalyzing the data the probabilities of the two categorieswere as follows P (U1) (512) and P (U2) (712) Basedon the formula the total entropy was as follows

I (U) minus512log2

512

minus712log2

712

09799 (7)

412 6e Conditional Entropy and Information Gain of EachInfluencing Factor In this study there were three factorsthat determined the purchase intentions of users A1 thestarting mode of the self-adaptive cruise A2 the position ofthe function key to adjust the distance between cars and A3the position of the function key of auto hold e proba-bilities of the factors and each condition are shown inTable 3

e information entropy values of the three factors wereas follows

E A1( 1113857 612

I(1 5) +612

I(4 2) 07842

E A2( 1113857 612

I(3 3) +612

I(2 4) 09592

E A3( 1113857 412

I(1 3) +412

I(1 3) +412

I(3 1) 08113

(8)

erefore the information gains of the factors were asfollows

Gain A1( 1113857 I(U) minus E A1( 1113857 01957

Gain A2( 1113857 I(U) minus E A2( 1113857 00207

Gain A3( 1113857 I(U) minus E A3( 1113857 01686

(9)

According to the information gains the starting mode ofthe self-adaptive cruise was the most important for the userrsquosperceived comfort and purchase intension while the posi-tion of the function key to adjust the distance between carswas not so significant By normalizing the information gainsof each influencing factor the results were taken as theweights of the factors which were

β1 Gain A1( 1113857

Gain A1( 1113857 + Gain A2( 1113857 + Gain A3( 1113857 05083

β2 Gain A2( 1113857

Gain A1( 1113857 + Gain A2( 1113857 + Gain A3( 1113857 00538

β3 Gain A3( 1113857

Gain A1( 1113857 + Gain A2( 1113857 + Gain A3( 1113857 04379

(10)

After taking these values into formula (1) the artificialfitness function could be constructed as follows

F 1

05083lowastQ1 + 00538lowastQ2 + 04379lowastQ3( 1113857 (11)

42 Results of GA

421 Characteristics of the Population e population sizeof this study was Pop 30 For the convenience of recordingthe unit of each dimension was decimeter (dm)e featuresof the 14 dimensions are described in Table 4

422 Solving Process of GA To match the userrsquos expecta-tion the seed of each initial population should be able to

Table 2 Classification of purchase intentions to interaction modes

e starting mode of self-adaptive cruise

e position of the function key to adjust thedistance between cars

e position of the function key ofauto-hold

Purchasedecision

Pulling the side stalk On the middle of the left spoke Under the EPB button 2Pressing the button On the middle of the left spoke Under the EPB button 2Pulling the side stalk On the bottom edge of the left spoke Under the EPB button 2Pressing the button On the bottom edge of the left spoke Under the EPB button 1Pulling the side stalk On the middle of the left spoke On the left of the EPB button 2Pressing the button On the middle of the left spoke On the left of the EPB button 1Pulling the side stalk On the bottom edge of the left spoke On the left of the EPB button 2Pressing the button On the bottom edge of the left spoke On the left of the EPB button 2Pulling the side stalk On the middle of the left spoke On the right of the EPB button 1Pressing the button On the middle of the left spoke On the right of the EPB button 1Pulling the side stalk On the bottom edge of the left spoke On the right of the EPB button 2Pressing the button On the bottom edge of the left spoke On the right of the EPB button 1

Table 3 e probabilities of the three factors and each condition

P (Vi) P (U1 | Vi) P (U2 | Vi)

A1Side stalk 612 16 56Button 612 46 26A2Middle 612 36 36Bottom edge 612 26 46A3Under the EPB button 412 14 34On the left of the EPB button 412 14 34On the right of the EPB button 412 34 14P (Vi) represents the probability of each condition of the influencingfactors P(U1 | Vi) and P(U2 | Vi) respectively indicate the conditionalprobabilities of the two categories

Complexity 7

cover part of the information of the PCD data And it shouldbe capable of satisfying the geometric description rules ofproduct evolution in terms of space and size [46] e initialpopulation of this paper came from repeated experiments inmany days and the perceptual experience and spatial at-tributes reflected by the data were stable e optimal so-lution of the model was obtained after 25 iterations Figure 3shows the evolutionary process of GA

In the figure the red line represents the best fitness valueof each generation And the blue dotted line describes themean fitness of each generation It could be seen from thefigure that the optimal value in the model increased with thenumber of iterations and converged into a straight line afterreaching the maximum value Until the optimal value of themodel remained unchanged the global optimal solutioncould be obtained

e genetic evolution stopped after 25 iterations It couldbe seen from the figure that although the deviation betweenthe optimal fitness values and the mean ones increased afterthe 9th iteration it continued to evolve which met theexpectation of evolution After the 21st iteration the devi-ation was close to 0

In the 25 generations of evolutionary iteration based onthe PCD data eight individuals with the highest fitness wereobtained which were the HMI dimension schemes that metthe usersrsquo expectation best in a certain generation of thepopulation It would be rational to make designs of HMIssuch as steering wheels and gear lever panels in the laterstage according to these fine values is could not onlyinherit the modelling genes of the HMIs but was alsocoupled to the user expectation to a certain extent e fi-nally generated dimensional data of the individuals with thehighest fitness are shown in Table 5

43 Significance of the Difference between the Original Di-mensions and the Evolved Ones To further validate whetherthe individuals with the highest fitness could bring about a

better take-over performance it was reasonable to collect theneeded time to complete driving assistant tasks under theHMI with the original sizes and the one with the evolvedsizes And the differences between the two sets of HMIdimensions could be analyzed

In the last of the 30 experiments we collected the time ofthe 13 subjects to complete the take-over tasks under theseven interaction modes in Table 1 Since the tasks had beenrepeated many times the subjects were skilled and the taskcompletion time would be relatively stable After making 3Dprinting models of the side stalk and the buttons accordingto the most suitable dimensions (Table 5) the drivingsimulator was reformeden the 13 subjects were invited torepeat the seven tasks with the new HMI and this set ofcompletion time was collected To judge whether the di-mensions could effectively improve the interaction perfor-mance and efficiency of driving information processingpaired-sample t-test was used on the two groups of data eresults are shown in Table 6

It could be seen from the results that except for A2 andC2 the completion time under the two sets of dimensions inthe process of the other five tasks all had a significant dif-ference (Figure 4)

Figure 4 shows that the completion time of each task wasshorter when the dimensions with high fitness were used onthe HMI is indicated that the driverrsquos information pro-cessing was faster which helped to improve the quality oftake-over and the ability to control the car As the driverrsquosattention to the HMI might compete with the execution ofdriving tasks for cognitive resources [47] and impair thedriving performance a shorter completion time of take-overtasks was conducive to the distribution of the driverrsquospsychological resources and more beneficial to trafficefficiency

It could be found by comparing the completion time ofeach task under the new HMI that (A) for starting the self-

Table 4 Features of the 14 dimensions

Tasks anddimensions

Mean(dm)

Stddeviation

Minimum(dm)

Maximum(dm)

A1 a11 19513 00789 18086 2091a12 06055 0063 05117 06977

A2 a21 08624 00242 08224 08979a22 00784 00477 00008 01487

B1 b11 13356 00633 12528 14475b12 00315 00169 00024 00588

B2 b21 13999 00526 13060 14952b22 0322 00139 03011 03497

C1 c11 29776 01254 28032 31940c12 3272 01414 30207 34928

C2 c21 27451 00305 27010 27966c22 32714 01304 30143 34819

C3 c31 32929 00596 32019 33963c32 3244 01523 30011 34932

2500735

0074

00745

0075

00755

0076

00765

0077

00775

0078

Fittn

ess

5 10 15 200Generation

Mean fitnessBest fitness

Figure 3 Solving process of GA

8 Complexity

adaptive cruise the completion time of operating with a sidestalk was shorter (B) for the task of adjusting the distancebetween cars the completion time would be shorter whenoperating with a button located in the middle of the steeringwheel spoke (C) for the task of auto hold a starting buttonon the left side of the EPB button could bring about a shortercompletion time ese results provided a reference for thefuture automobile interior design

5 Discussion

Driving assistant system (DAS) plays an important role inthe context of human-computer codriving As reportedDAS can help the driver to complete the intense drivingtasks effectively and abate the pressure level of the driver[48] erefore it is meaningful to make a study for thecontroller dimensions in the DAS e study collectedcomfortable dimensions perceived by the drivers in a relaxedstate e aim was to detect a set of reasonable dimensions

which could make drivers at a low level of stress In ad-dition it is valuable for drivers to adapt to the different roadenvironment and realize sustainable traffic to optimize thesizes of interactive components on the in-vehicle HMI emain goals of in-vehicle technologies and cooperativeservices were to reduce traffic congestion and improve roadsafety [4]

e study attempted to analyze the optimal HMI di-mensions suitable for take-over tasks e involved HMIcomponents such as the steering wheel self-adaptive cruiselever distance adjustment button and auto hold buttonwere all direct design elements for users to see and touchBecause the design and performance of a product wereimportant factors affecting the decision-making of con-sumers [49] we calculated the information entropy andinformation gains of the three kinds of take-over tasks forclassifying usersrsquo purchase intentions e informationtransmitted by a product was multivariant complex andfuzzy so that there would be dissipation in the process oftransmission erefore the image evaluation process ofproducts could be regarded as a chaotic state e feedbackinformation could be calculated by information entropy tocarry out a comprehensive evaluation e smaller the en-tropy of an image the clearer it was From the perspective ofchaos degree the 2nd factor in Table 3 (the position of thefunction key to adjust the distance between cars) broughtmore uncertainty to the purchase decision(E(A2) 09592) which indicated that little informationwas provided by the 2nd factor for judging the usersrsquo de-cisions On the contrary the uncertainty produced by the 1stfactor (the starting mode of the self-adaptive cruise) was thesmallest (E(A1) 07842) is illustrated that using theside stalk or a button to complete the switch of the controlright had a great impact on the satisfaction and purchasedecision of users

After normalizing the information entropy of threekinds of take-over tasks their weight values were calculated

Table 5 e finally generated dimensional data of the individuals with the highest fitness

(A) e starting mode of self-adaptive cruise

(B) e position of thefunction key to adjust thedistance between cars

(C) e position of the function key of auto-hold Fitness value

a11 a12 a21 a22 b11 b12 b21 b22 c11 c12 c21 c22 c31 c32202 062 089 009 136 004 148 030 284 302 271 338 321 313 008

Table 6 e paired sample t-test results of task completion time under the two HMIs with different dimensions

Interaction modes of take-over tasks

e task completiontime under the original

dimensions (s)

e task completiontime under the evolved

dimensions (s) t-value Sig

Mean Std dev Mean Std devA1lowast 0715 0185 0537 0167 2553 0025A2 0657 0181 0637 0186 1151 0272B1lowast 0648 013 0462 0206 2797 0016B2lowastlowast 0736 0106 0571 0124 3116 0009C1lowast 0753 0295 0562 0222 2515 0027C2 0667 0226 0522 0176 1999 0069C3lowast 069 0208 0537 0225 2355 0036

A1lowast B1lowast B2lowastlowast C1lowast C3lowast

0

02

04

06

08

1

12

Time under the original dimensionsTime under the evolved dimensions

Figure 4 e completion time of the tasks with significant dif-ferences under the two HMIs with different dimensions

Complexity 9

and the fitness function was constructed In this way thedimensions with the highest fitness could be found out bymeans of GA

e background and driving experience of the subjects weredifferent and their familiarity with the take-over tasks and theHMI was variant Besides the noise of usersrsquo subjective per-ception data was always high Taking the above factors intoconsideration we collected PCD data repeatedly from the same13 subjects to form the initial population From the evolutionarytrajectory it could be found that in the 8th 10th 11th 14th 17thand 21st iteration there appeared an optimal solution

GA was always used for parameter optimization bysearching the optimal solutions while seldom used in userresearch and environment perception analysis But in studiesof user-centred design it could search the set of satisfactorysolutions driven by user expectations [46] Many pieces ofresearch had utilized GA to improve the ergonomic effi-ciency or optimize interactive component sizes in an op-erating environment [34ndash38 50 51] In this study wereported a driver perception research which investigated themost comfortable HMI sizes under different interactionmodes of take-over tasks By means of normalized geometricselection arithmetic uniform crossover and nonuniformmutation GA was applied in the research of perceivedcomfortable dimensions and a set of reasonable dimensionswas obtained e dimensional evolutionary method dis-cussed in this study embodied the concept of dynamic in-teraction design which took the mutual influence andrestrictions among different dimensions into considerationCompared with traditional methods of measuring manytimes and taking the average this method was more ad-vanced And the study enriched theoretical researches in thefield of product design and collaborative optimization

Perceived comfort was a complex problem for its dif-ficulty to be quantified objectively But it was useful in er-gonomic explorations Researchers had found that theperceptual messages from product property and usage re-quirements were competent to the analysis of user experi-ence [52] erefore it was reasonable to probe suitabledimensions by perceived comfort For the study of HMIdesign effect in a specific environment these data couldproduce important and dependable results [53] Rationaldimensions of the automobile interior HMI played an im-portant role in the driverrsquos operation quality Caruso [54]studied driversrsquo experience on the seat by an experiment andfound that interior space influenced the driversrsquo ergonomicperformance [54] For the take-over tasks in the situation ofhuman-computer codriving the dimensions also affectedwhether the driver could safely achieve the control rightswitch in the shortest possible time

In this paper paired-sample t-test was used to testwhether the sizes with the highest fitness could bring about asignificant reduction in task completion time In the resultswe found that when the interface adopted the finally evolvedsizes the completion time of each task was shorter As thecompletion time of in-vehicle interactive tasks was related tothe vehicle speed and driver errors under the context ofhuman-computer codriving [21 55] a shorter completiontime would help the driver to control the speed better

6 Limitation

e study explored the reasonable dimensions of the hu-man-machine interface in intelligent cars Neverthelessthere were still some deficiencies in the equipment used inthis study e driving simulator was easier to control andreform and helped to reduce the subjectsrsquo pressure How-ever the immersion visual interactivity and fidelity degreewere not high enough After obtaining reliable conclusionsin this study outdoor experiment with a real vehicle could beconducted in the next stage

Besides the sample size of this study was limited Whenthe sample size was larger in future studies the represen-tativeness of the data would be better and the results of GAcould be further improved

Finally usersrsquo perception of the HMI dimensions de-viated from each other In the next phase of the studyelectroencephalography (EEG) signal in the process ofperforming the driving tasks could be used By this meansthe subjectsrsquo attention and intention when interacting withthe HMI could be reflected in an objective way [56 57] as asupplement to their subjective report

7 Conclusion

For HMI design of industrial products physical concretemodelling parameters such as the product shape size andlayout were the objective properties that users could directlyperceive and the basis for the product to be expressed Basedon the dynamic measurement principle this paper proposedan optimized design method for ergonomic dimensions byGA e results indicated that the method was effective andprovided a new way for product shape innovation andhuman-machine dimensional automatic design e mainconclusions of this study were as follows

(1) Based on the information entropy theory the weightvalues of the three driving assistance functions werecalculated e results showed that the informationgain of the starting mode of the self-adaptive cruisewas highest while the information gain of the po-sition of the function key to adjust the distancebetween cars was lowest which provided a theo-retical basis for the dimensional fitness evaluationmechanism of the HMI

(2) e evolutionary process of HMI dimensionsdriven by user expectation was studied and theevolution conditions were analyzed e studyconfirmed the feasibility of normalized geometricselection arithmetic uniform crossover and non-uniform mutation in iterative researches for hu-man-machine dimensions rough parameterdebugging this evolutionary method was proved tobe effective in obtaining appropriate sizes of HMIcomponents e model could make full use of theadvantages of GA to improve the user researchperformance such as few restrictions on the opti-mization high robustness and an easily achievedglobal search

10 Complexity

(3) After finding the dimensions with the highest fitnessand applying them to the design of the HMI within-subjects experiments showed that the completiontime of five take-over tasks was significantly shorterexcept for that of Task A2 and C2is indicated thatthe information processing of drivers under the HMIwith evolved dimensions was faster which was morebeneficial to the vehicle control and take-overquality Intelligent vehicles achieved intelligent in-formation exchange with people vehicles and roadsthrough in-vehicle sensor system and informationterminals erefore the evolved dimensions couldmake the vehicle apt to respond to the environmentand have more time to fully deal with the complextraffic environment and would help to improve thetraffic efficiency therefrom

Data Availability

e data used to support the findings of this study areavailable from the corresponding author upon request

Consent

e study respected and guaranteed the subjectsrsquo right toparticipate in the study and allowed them to withdraw from thestudy unconditionally at any stage e study strictly followedthe informed consent procedure e subjects were asked tosign a consent form to inform them of the purpose processduration of the experiment matters needing attention and themethods and purpose of collecting the data so as to reducetheir anxiety and confusion In the experiment the personalsafety and health of the subjects took priority According to theusage guidelines of the driving simulator we tried to avoid anypossible injury in driving tasks e privacy of the subjects wasstrictly protected e subjects were informed of the storageuse and confidentiality measures of their personal informationand image data in advance truthfully And we would notdisclose the information to a third party or spread them on theInternet without authorization

Conflicts of Interest

e authors declare that they have no conflicts of interest

Acknowledgments

is researchwas funded by BeijingUrbanGovernance ResearchProject (grant no 20XN244) Scientific Research Program ofBeijing Education Commission (grant no KM202010009003)Yuyou Talent Support Program of North China University ofTechnology (grant no 107051360018XN012018) Tianjin YouthTalents Reserve Support Program and Chinese ErgonomicsSociety amp Kingfar Joint Research Fund for Outstanding YoungScholars (grant number CES-Kingfar-2019-001)

References

[1] S Jin J Yang E Wang and J Liu ldquoe influence of high-speed rail on ice-snow tourism in north eastern ChinardquoTourism Management vol 78 Article ID 104070 2020

[2] J Yang R Yang J Sun T Huang and Q Ge ldquoe spatialdifferentiation of the suitability of ice-snow tourist destina-tions based on a comprehensive evaluation model in ChinardquoSustainability vol 9 no 5 Article ID 774 2017

[3] W-Y Chung T-W Chong and B-G Lee ldquoMethods to detectand reduce driver stress a reviewrdquo International Journal ofAutomotive Technology vol 20 no 5 pp 1051ndash1063 2019

[4] H Farah and H N Koutsopoulos ldquoDo cooperative systemsmake driversrsquo car-following behavior saferrdquo TransportationResearch Part C Emerging Technologies vol 41 pp 61ndash722014

[5] SAE Taxonomy and Definitions for Terms Related to DrivingAutomation Systems for On-Road Motor Vehicles SAEJ3016TM Warrendale PA USA 2018

[6] J Hoegg J W Alba and D W Dahl ldquoe good the bad andthe ugly influence of aesthetics on product feature judg-mentsrdquo Journal of Consumer Psychology vol 20 no 4pp 419ndash430 2010

[7] M Kyriakidis R Happee and J C F De Winter ldquoPublicopinion on automated driving results of an internationalquestionnaire among 5000 respondentsrdquo TransportationResearch Part F Traffic Psychology and Behaviour vol 32pp 127ndash140 2015

[8] F Meng and C Spence ldquoTactile warning signals for in-vehiclesystemsrdquo Accident Analysis amp Prevention vol 75 pp 333ndash346 2015

[9] J Wan and C Wu ldquoe effects of lead time of take-overrequest and nondriving tasks on taking-over control of au-tomated vehiclesrdquo IEEE Transactions on Human MachineSystems vol 48 no 6 pp 582ndash591 2018

[10] S Petermeijer P Hornberger I Ganotis J D Winter andK Bengler ldquoe design of a vibrotactile seat for conveyingtake-over requests in automated drivingrdquo in Proceedings of the8th International Conference on Applied Human Factors andErgonomics Los Angeles CA USA July 2017

[11] T Ito A Takata and K Oosawa ldquoTime required for take-overfrom automated to manual drivingrdquo in Proceedings of the SAE2016 World Congress and Exhibition Detroit MI USA April2016

[12] K Zeeb A Buchner and M Schrauf ldquoWhat determines thetake-over time An integrated model approach of driver take-over after automated drivingrdquo Accident Analysis amp Preven-tion vol 78 pp 212ndash221 2015

[13] B Mok M Johns K J Lee et al ldquoEmergency automation offunstructured transition timing for distracted drivers of au-tomated vehiclesrdquo in Proceedings of the IEEE InternationalConference on Intelligent Transportation Systems Las PalmasSpain September 2015

[14] S Samuel A Borowsky S Zilberstein and D L FisherldquoMinimum time to situation awareness in scenarios involvingtransfer of control from an automated driving suiterdquoTransportation Research Record Journal of the TransportationResearch Board vol 2602 no 2602 pp 115ndash120 2016

[15] W Vlakveld N van Nes J de Bruin L Vissers andM van der Kroft ldquoSituation awareness increases when drivershave more time to take over the wheel in a level 3 automated

Complexity 11

car a simulator studyrdquo Transportation Research Part F TrafficPsychology and Behaviour vol 58 pp 917ndash929 2018

[16] N Merat A H Jamson F C H Lai M Daly andO M J Carsten ldquoTransition to manual driver behaviourwhen resuming control from a highly automated vehiclerdquoTransportation Research Part F Traffic Psychology and Be-haviour vol 27 no 26 pp 274ndash282 2014

[17] C Z Wu H R Wu and N C Lyu ldquoReview of control switchand safety of human-computer driving intelligent vehiclerdquoJournal of Traffic and Transportation Engineering vol 18no 6 pp 131ndash141 2018

[18] C Gold D Dambock L Lorenz and K Bengler ldquoldquoTakeoverrdquo how long does it take to get the driver back into thelooprdquo Proceedings of the Human Factors and ErgonomicsSociety Annual Meeting vol 57 no 1 pp 1938ndash1942 2013

[19] H Yang Y Zhao and Y Wang ldquoIdentifying modeling formsof instrument panel system in intelligent shared cars a studyfor perceptual preference and in-vehicle behaviorsrdquo Envi-ronmental Science and Pollution Research vol 27 no 1pp 1009ndash1023 2020

[20] A Berkovich A Lu B Levine and A V Reddy ldquoObservedcustomer seating and standing behavior and seat preferenceson board subway cars in New York cityrdquo TransportationResearch Record Journal of the Transportation ResearchBoard vol 2353 no 1 pp 33ndash46 2013

[21] R Li Y V Chen C Sha and Z Lu ldquoEffects of interface layouton the usability of in-vehicle information systems and drivingsafetyrdquo Displays vol 49 pp 124ndash132 2017

[22] H Kim S Kwon J Heo H Lee and M K Chung ldquoe effectof touch-key size on the usability of in-vehicle informationsystems and driving safety during simulated drivingrdquo AppliedErgonomics vol 45 no 3 pp 379ndash388 2014

[23] H Kim and H Song ldquoEvaluation of the safety and usability oftouch gestures in operating in-vehicle information systemswith visual occlusionrdquo Applied Ergonomics vol 45 no 3pp 789ndash798 2014

[24] M Koerber C Gold D Lechner and K Bengler ldquoe in-fluence of age on the take-over of vehicle control in highlyautomated drivingrdquo Transportation Research Part F TrafficPsychology amp Behaviour vol 39 pp 19ndash32 2016

[25] H Clark A C Mclaughlin B Williams and J Feng ldquoPer-formance in takeover and characteristics of non-driving re-lated tasks during highly automated driving in younger andolder driversrdquo Proceedings of the Human Factors amp Ergo-nomics Society Annual Meeting vol 61 no 1 pp 37ndash41 2017

[26] I Lijarcio S A Useche J Llamazares and L MontoroldquoPerceived benefits and constraints in vehicle automationdata to assess the relationship between driverrsquos features andtheir attitudes towards autonomous vehiclesrdquo Data in Briefvol 27 Article ID 104662 2019

[27] K-Z Chen and X-A Feng ldquoVirtual genes of manufacturingproducts and their reforms for product innovative designrdquoProceedings of the Institution of Mechanical Engineers Part CJournal of Mechanical Engineering Science vol 218 no 5pp 557ndash574 2004

[28] K-Z Chen X-A Feng and X-C Chen ldquoReverse deductionof virtual chromosomes of manufactured products for theirgene-engineering-based innovative designrdquo Computer-AidedDesign vol 37 no 11 pp 1191ndash1203 2005

[29] H-C Tsai and J-R Chou ldquoAutomatic design support andimage evaluation of two-coloured products using colour as-sociation and colour harmony scales and genetic algorithmrdquoComputer-Aided Design vol 39 no 9 pp 818ndash828 2007

[30] S-W Hsiao C-F Hsu and K-W Tang ldquoA consultation andsimulation system for product color planning based on in-teractive genetic algorithmsrdquo Color Research amp Applicationvol 38 no 5 pp 375ndash390 2013

[31] Y Sun W-l Wang Y-w Zhao and X-j Liu ldquoUser imageoriented interactive genetic algorithm evaluation mode inproduct developmentrdquo Computer Integrated ManufacturingSystems vol 18 no 2 pp 276ndash281 2012

[32] R Kamalian E Yeh Z Ying A M Agogino and H TakagildquoReducing human fatigue in interactive evolutionary com-putation through fuzzy systems and machine learning sys-temsrdquo in Proceedings of the 2006 IEEE InternationalConference on Fuzzy Systems Vancouver Canada July 2006

[33] Z Gu M Xi Tang and J H Frazer ldquoCapturing aestheticintention during interactive evolutionrdquo Computer-AidedDesign vol 38 no 3 pp 224ndash237 2006

[34] G Harih M Borovinsek and Z Ren Optimisation ofProductrsquos Hand-Handle Interface Material Parameters forImproved Ergonomics Springer International PublishingCham Switzerland 2015

[35] P Cheng D Chen and J Wang ldquoClustering of the bodyshape of the adult male by using principal component analysisand genetic algorithm-BP neural networkrdquo Soft Computingvol 24 no 17 Article ID 13219 2020

[36] N M B Serrano P J G Nieto A S Sanchez F S Lasherasand P R Fernandez AHybrid Algorithm for the Assessment ofthe Influence of Risk Factors in the Development of Upper LimbMusculoskeletal Disorders Springer Cham Switzerland 2018

[37] S S Sankar O-M Holman A F Gazabon and C J AcevedoldquoApplication of genetic algorithm to job scheduling underergonomic constraints in manufacturing industryrdquo Journal ofAmbient Intelligence and Humanized Computing vol 10no 5 pp 2063ndash2090 2019

[38] A-Z Atiya L Lee and K Xing ldquoDeveloping a multi-ob-jective genetic optimisation approach for an operationaldesign of a manual mixed-model assembly line with walkingworkersrdquo Journal of Intelligent Manufacturing vol 27 no 5pp 1ndash17 2014

[39] P Jedlicka and T Ryba ldquoGenetic algorithm application inimage segmentationrdquo Pattern Recognition and Image Anal-ysis vol 26 no 3 pp 497ndash501 2016

[40] L V Campenhout J Frens K Overbeeke A Standaert andH Peremans ldquoPhysical interaction in a dematerializedworldrdquo International Journal of Design vol 7 no 1 pp 1ndash182013

[41] K Nandhini and S R Balasundaram ldquoImproving readabilitythrough individualized summary extraction using interactivegenetic algorithmrdquo Applied Artificial Intelligence vol 30no 7 pp 635ndash661 2016

[42] R M Gray Entropy and Information 6eory Springer NewYork NY USA 2nd edition 2011

[43] B Anderhofstadt and S Spinler ldquoFactors affecting the pur-chasing decision and operation of alternative fuel-poweredheavy-duty trucks in Germanymdasha Delphi studyrdquo Trans-portation Research Part D Transport and Environmentvol 73 pp 87ndash107 2019

[44] X-m Wu and Y Feng ldquoReactive power optimization ofdistribution network based on improved genetic algorithmrdquoJournal of Xirsquoan Shiyou University (Natural Science Edition)vol 30 no 3 pp 95ndash99 2015

[45] Z Ling andM-j Li ldquoState space evolutionary algorithm basedon non-uniform mutation operatorrdquo Computer Technologyand Development vol 28 no 9 pp 68ndash71 2018

12 Complexity

[46] W Hu and J Zhao ldquoAutomobile styling gene evolutiondriven by usersrsquo expectation imagerdquo Journal of MechanicalEngineering vol 47 no 16 pp 176ndash181 2011

[47] J J Scott and R Gray ldquoA comparison of tactile visual andauditory warnings for rear-end collision prevention in sim-ulated drivingrdquo Human Factors 6e Journal of the HumanFactors and Ergonomics Society vol 50 no 2 pp 264ndash2752008

[48] B Reimer B Mehler and J F Coughlin ldquoReductions in self-reported stress and anticipatory heart rate with the use of asemi-automated parallel parking systemrdquo Applied Ergonom-ics vol 52 pp 120ndash127 2016

[49] B DrsquoIppolito ldquoe importance of design for firms com-petitiveness a review of the literaturerdquo Technovation vol 34no 11 pp 716ndash730 2014

[50] S Asensio-Cuesta J A Diego-Mas L Canos-Daros andC Andres-Romano ldquoA genetic algorithm for the design of jobrotation schedules considering ergonomic and competencecriteriardquo International Journal of Advanced ManufacturingTechnology vol 60 no 9ndash12 pp 1161ndash1174 2012

[51] S-C Lee H-E Tseng C-C Chang and Y-M HuangldquoApplying interactive genetic algorithms to disassembly se-quence planningrdquo International Journal of Precision Engi-neering and Manufacturing vol 21 no 4 pp 663ndash679 2020

[52] M G Shishaev V V Dikovitsky and L V Lapochkina ldquoeexperience of building cognitive user interfaces of multido-main information systems based on the mental model ofusersrdquo in Proceedings of the 6th Computer Science On-LineConference (CSOC 2017) Prague Czech Republic April 2017

[53] S Lee H Alzoubi and S Kim ldquoe effect of interior designelements and lighting layouts on prospective occupantsrsquoperceptions of amenity and efficiency in living roomsrdquo Sus-tainability vol 9 no 7 Article ID 1119 2017

[54] G Caruso ldquoMixed reality system for ergonomic assessment ofdriverrsquos seatrdquo International Journal of Virtual Reality vol 10no 2 pp 69ndash79 2011

[55] M L Reyes and J D Lee ldquoEffects of cognitive load presenceand duration on driver eye movements and event detectionperformancerdquo Transportation Research Part F Traffic Psy-chology and Behaviour vol 11 no 6 pp 391ndash402 2008

[56] D Cisler P M Greenwood D M Roberts R McKendrickand C L Baldwin ldquoComparing the relative strengths of EEGand low-cost physiological devices in modeling attentionallocation in semiautonomous vehiclesrdquo Frontiers in HumanNeuroscience vol 13 Article ID 109 2019

[57] I-H Kim J-W Kim S Haufe and S-W Lee ldquoDetection ofbraking intention in diverse situations during simulateddriving based on EEG feature combinationrdquo Journal of NeuralEngineering vol 12 no 1 Article ID 016001 2014

Complexity 13

of the perceptual experience the experiment wouldbe in progress for six days As a result each subjectwould carry out the tasks for 5lowast 6 30 times ePCD data collected each time were taken as the initialpopulation and the dimensions with the highestfitness would be obtained by GA

(8) In the 30th experiment the completion time anderror times of the subjects during the seven take-overtasks were collected

(9) After obtaining the PCD data with the highestadaptability models of the side stalk and buttonswere made by 3D printers to replace the originalones e subjects were asked to repeat the seventasks in Figure 2 again with the modified simulatorAnd their completion time and error times under thenew HMI would be collected

32 Information Entropy and Genetic Algorithm GA is aself-adaptive global optimization algorithm which imitatesthe natural selection and individual heredity of the biologicalworld [39] Traditionally interactive genetic algorithm(IGA) was widely used in product modelling design forsubject recognition [40] However IGA has some short-comings including the uncertainty of individual fitnessvalues the nonpersistence of the individual evaluationprocess and the nonuniqueness of optimization results [41]In this study GA was taken as the main research methodwhile PSC was used in the construction of fitness functionwhich integrated user preference intuition emotion andpsychological optimization into it erefore the fitnessvalues of evolutionary individuals could make the pop-ulation evolve in the direction that users expected

321 Fitness Function Whether the population can evolvein the direction of the ideal solution depends on the es-tablishment of fitness function to a great extent Accordingto the userrsquos PSC and the weight of the three DAFs thatneeded the driver to take over the artificial fitness functionwas formulated as follows

F 1

β1Q1 + β2Q2 + β3Q3( 1113857 (1)

Among which β1 sim β3 represented the importanceweights of the three DAFs which were obtained by calcu-lating the information entropy Q1 sim Q3 meant the com-fortable dimension of the operating components of eachDAF listed in Table 1 Q1 a11 lowast a12 + a21 lowast a22 Q2 b11 lowastb12 + b21 lowast b22 Q3 c11 lowast c12 + c21 lowast c22+ c31 lowast c32 emeanings of a11 sim c32 values are shown in Table 1 whichwere obtained in step (7) of the experiment For each in-teraction mode the comfortable position of the corre-sponding operation component was identified with a pointat first and then the distances between the point and thehorizontal and vertical centerlines of the steering wheel weremeasurede two distances were multiplied to get a relativecomfort area F was used to evaluate the fitness degree of

individuals With a smaller size the userrsquos behavioral path tocomplete the operations could be shorter and the taskcompletion time would be relatively less which were con-ducive to the completion of the take-over tasks ereforethe larger the F value was the more reasonable the sizeswere

322 Information Entropy and the Weight of the DrivingAssistance Functions Based on the entropy theory the studyexplored the importance level of the main DAFs in L1-L2automated driving and analyzed the influencing factors ofusersrsquo purchase intentions e uncertainty degree of thedivided sample set which was measured by means of cal-culating the information gain was used as the standard toweigh the quality of the division e larger the informationgain the less the uncertainty degree of the sample set

Entropy refers to the degree of system chaos which is ameasurement of the possibility of the system in a certainmacroscopic state Claude Elwood Shannon put forward theconcept of information entropy to express the order degreeof a system [42] Let S be a set of s samples Suppose theclassification attribute has m different values Ci(i

1 2 m) and let si be the number of samples in class Cien for a given sample its total entropy isI(s1 s2 sm) minus 1113936

mi1 Pilog2(Pi) Among which Pi is the

probability that any sample belongs to CiLet an attributeA have k different values a1 a2 ak1113864 1113865

Using the attribute to divide the set S into k subsetsS1 S2 Sk1113864 1113865 Among which Sj contains the samples with avalue aj in the set S Let sij be the number of samples of classCj in subset Sjen the information entropy of the samplesdivided according to A is given by

E(A) 1113944k

j1

s1j + s2j + middot middot middot + smj

sI s1j s2j smj1113872 1113873 (2)

where I(s1j s2j smj) minus 1113936mi1 Pijlog2(Pij) And Pij

((s1j + s2j + middot middot middot + smj)s) is the probability of the samples ofclass Cj in subset Sj

Finally the information gain obtained by dividing the setS according to the attribute A is Gain(A)

I(s1 s2 sm) minus E(A) Obviously the smaller the E(A)the larger the Gain(A) which means more information isprovided by A to judge the usersrsquo purchase intention and theimportance of A is higher After normalizing the Gain(A)

values of the three DAFs the weight values β1 sim β3 in thefitness function could be formed

We sent questionnaires online to eight professional usersto investigate their purchase decisions of products formed bydifferent combinations of interaction modes e purchasedecisions were divided into two categories which were ldquobuyrdquoand ldquonot buyrdquo Delphi method was used to collect theiropinions in an all-round way and ensure the consistency oftheir decisions Compared with taking the most frequentlyselected option as the final decision the results of the Delphimethod were more scientific and more information wasavailable in the research [43]

Complexity 5

323 Genetic Operation and Evolutionary Process In thissystem the premise to debug the genetic operation andmakejudgment was due to the irrationality and uncertainty of theshape and spatial dimension design of automobile con-trollers and panels the dimension evolution driven by usersrsquoexpectation was not only to search for the optimal solutionbut also to obtain the satisfactory solution under limitedresources erefore the model was different in codingselection crossover and mutation from traditional GAwhich was to solve problems represented by explicitfunctions

(1) Coding Generally there are two kinds of codingmethods for optimization problems which are real numbercoding and binary coding And both have advantages anddisadvantages Mapping errors exist in traditional binarycoding when continuous functions are discretizedWhen theindividual coding string is short the accuracy requirementsmay not be met While when the string is long although thecoding accuracy can be improved the search space of GAwill expand dramatically [44] erefore we used realnumber coding in this study

In the process of evolution a population Pop(t) wasformed Each chromosome Xt

i in the population representeddata from one of the 30 experiments e genes in thechromosome were composed of PCD reported by thesubjects after completing the whole experiment processwhich could be expressed as follows

Pop(t) Xt1 X

t2 X

ti X

tM1113966 1113967

Xti θt

i1 θti2 θt

ib θtiN1113960 1113961

(3)

where t represented the generation number of evolutions θrepresented gene coding and M represented the populationsize

(2) Initialized Population and Selection e populationneeded to be initialized before running GA According to theresearch demand PCD data from the 30 experiments weretaken as the primary population Each time data of the 14PCD indexes were collected to form the population matrix

e selection operation was the process of selectingindividuals from the previous generation of the populationto form the next generation e purpose of selection was toobtain fine individuals based on the fitness values so thatthey would have the opportunity to reproduce as parents forthe next generation Individuals with high fitness were morelikely to be inherited while the ones with low fitness wereless likely After debugging and comparison we chose themethod of normalized geometric select (NGS) to performthe selection operation NGS was mainly to sort the fitnessvalues and the better individuals would be maintained asparents is was more suitable for the selection of productdesign schemes and also helpful to prevent better individualsfrom being damaged

(3) Crossover and Mutation By crossover operation anew generation of individuals could be obtained whichcombined the characteristics of their parents and embodiedthe idea of information exchange In the case of real number

coding the arithmetic uniform crossover (AUC) betweenindividuals was generally utilized [45] AUC was a linearcombination of two individuals to produce two new onesemethod of crossover and getting the next generation wasas follows

c1 p1lowast a + p2lowast (1 minus a)

c2 p1lowast (1 minus a) + p2lowast a(4)

where p1 and p2 were the parents and a was a random mixamount

In this study the nonuniform mutation (NM) was usedto perform the mutation operation Let an individual beX X1X2 Xk Xl If Xk was the variation point andits value range was [Uk

min Ukmax] a new individual

X X1X2 Xkprime Xl could be obtained after non-

uniformly mutating at this point And the new gene valuewas given by

Xkprime

Xk + Δ t Ukmax minus Xk1113872 1113873 if random(0 1) 0

Xk minus Δ t Xk minus Ukmin1113872 1113873 if random(0 1) 1

⎧⎨

⎩

(5)

where Δ(t y) was a random number conforming to thenonuniform distribution in the range [0 y] y representedUk

max minus Xk and Xk minus Ukmin It was required that with the

increase of evolutionary generation number t the proba-bility of Δ(t y) approaching to 0 also raised gradually

In this study Δ(t y) was defined as follows

Δ(t y) y 1 minus r(1minus tT)b

1113872 1113873 (6)

where r was a random number conforming to the non-uniform distribution in the range [0 1] which was theselection pressure T was the maximum number of evolu-tionary generations in this study T 25 In the 25 iterationsthe better individuals with high fitness could be found out toexplore the evolutionary mechanism of the appropriate HMIdimensions of the DAFs b was a parameter to adjust thevariable step size which was a system parameter It deter-mined the dependence degree of the random number dis-turbance on t and the range was usually 2sim5 Afterdebugging the model and making comparisons repeatedlythis parameter was set as b 4 in this study

4 Results

41 Information Entropy and the Weights of the DAFs

4116e Expectation Information for the Classification of theHMI of Take-Over Tasks According to different combina-tions of the seven interaction modes in Table 1 a total of2lowast 2lowast 312 HMI schemes could be generated By means ofDelphi method purchase decisions for the 12 schemes wereinvestigated After four rounds of expert investigations thepurchase decisions of the eight professional users reached anagreement e classification results are shown in Table 2

According to the purchase intentions of the professionalusers 12 HMI schemes could be divided into two categories

6 Complexity

U1 1 (buy) and U2 2 (not buy) By summarizing andanalyzing the data the probabilities of the two categorieswere as follows P (U1) (512) and P (U2) (712) Basedon the formula the total entropy was as follows

I (U) minus512log2

512

minus712log2

712

09799 (7)

412 6e Conditional Entropy and Information Gain of EachInfluencing Factor In this study there were three factorsthat determined the purchase intentions of users A1 thestarting mode of the self-adaptive cruise A2 the position ofthe function key to adjust the distance between cars and A3the position of the function key of auto hold e proba-bilities of the factors and each condition are shown inTable 3

e information entropy values of the three factors wereas follows

E A1( 1113857 612

I(1 5) +612

I(4 2) 07842

E A2( 1113857 612

I(3 3) +612

I(2 4) 09592

E A3( 1113857 412

I(1 3) +412

I(1 3) +412

I(3 1) 08113

(8)

erefore the information gains of the factors were asfollows

Gain A1( 1113857 I(U) minus E A1( 1113857 01957

Gain A2( 1113857 I(U) minus E A2( 1113857 00207

Gain A3( 1113857 I(U) minus E A3( 1113857 01686

(9)

According to the information gains the starting mode ofthe self-adaptive cruise was the most important for the userrsquosperceived comfort and purchase intension while the posi-tion of the function key to adjust the distance between carswas not so significant By normalizing the information gainsof each influencing factor the results were taken as theweights of the factors which were

β1 Gain A1( 1113857

Gain A1( 1113857 + Gain A2( 1113857 + Gain A3( 1113857 05083

β2 Gain A2( 1113857

Gain A1( 1113857 + Gain A2( 1113857 + Gain A3( 1113857 00538

β3 Gain A3( 1113857

Gain A1( 1113857 + Gain A2( 1113857 + Gain A3( 1113857 04379

(10)

After taking these values into formula (1) the artificialfitness function could be constructed as follows

F 1

05083lowastQ1 + 00538lowastQ2 + 04379lowastQ3( 1113857 (11)

42 Results of GA

421 Characteristics of the Population e population sizeof this study was Pop 30 For the convenience of recordingthe unit of each dimension was decimeter (dm)e featuresof the 14 dimensions are described in Table 4

422 Solving Process of GA To match the userrsquos expecta-tion the seed of each initial population should be able to

Table 2 Classification of purchase intentions to interaction modes

e starting mode of self-adaptive cruise

e position of the function key to adjust thedistance between cars

e position of the function key ofauto-hold

Purchasedecision

Pulling the side stalk On the middle of the left spoke Under the EPB button 2Pressing the button On the middle of the left spoke Under the EPB button 2Pulling the side stalk On the bottom edge of the left spoke Under the EPB button 2Pressing the button On the bottom edge of the left spoke Under the EPB button 1Pulling the side stalk On the middle of the left spoke On the left of the EPB button 2Pressing the button On the middle of the left spoke On the left of the EPB button 1Pulling the side stalk On the bottom edge of the left spoke On the left of the EPB button 2Pressing the button On the bottom edge of the left spoke On the left of the EPB button 2Pulling the side stalk On the middle of the left spoke On the right of the EPB button 1Pressing the button On the middle of the left spoke On the right of the EPB button 1Pulling the side stalk On the bottom edge of the left spoke On the right of the EPB button 2Pressing the button On the bottom edge of the left spoke On the right of the EPB button 1

Table 3 e probabilities of the three factors and each condition

P (Vi) P (U1 | Vi) P (U2 | Vi)

A1Side stalk 612 16 56Button 612 46 26A2Middle 612 36 36Bottom edge 612 26 46A3Under the EPB button 412 14 34On the left of the EPB button 412 14 34On the right of the EPB button 412 34 14P (Vi) represents the probability of each condition of the influencingfactors P(U1 | Vi) and P(U2 | Vi) respectively indicate the conditionalprobabilities of the two categories

Complexity 7

cover part of the information of the PCD data And it shouldbe capable of satisfying the geometric description rules ofproduct evolution in terms of space and size [46] e initialpopulation of this paper came from repeated experiments inmany days and the perceptual experience and spatial at-tributes reflected by the data were stable e optimal so-lution of the model was obtained after 25 iterations Figure 3shows the evolutionary process of GA

In the figure the red line represents the best fitness valueof each generation And the blue dotted line describes themean fitness of each generation It could be seen from thefigure that the optimal value in the model increased with thenumber of iterations and converged into a straight line afterreaching the maximum value Until the optimal value of themodel remained unchanged the global optimal solutioncould be obtained

e genetic evolution stopped after 25 iterations It couldbe seen from the figure that although the deviation betweenthe optimal fitness values and the mean ones increased afterthe 9th iteration it continued to evolve which met theexpectation of evolution After the 21st iteration the devi-ation was close to 0

In the 25 generations of evolutionary iteration based onthe PCD data eight individuals with the highest fitness wereobtained which were the HMI dimension schemes that metthe usersrsquo expectation best in a certain generation of thepopulation It would be rational to make designs of HMIssuch as steering wheels and gear lever panels in the laterstage according to these fine values is could not onlyinherit the modelling genes of the HMIs but was alsocoupled to the user expectation to a certain extent e fi-nally generated dimensional data of the individuals with thehighest fitness are shown in Table 5

43 Significance of the Difference between the Original Di-mensions and the Evolved Ones To further validate whetherthe individuals with the highest fitness could bring about a

better take-over performance it was reasonable to collect theneeded time to complete driving assistant tasks under theHMI with the original sizes and the one with the evolvedsizes And the differences between the two sets of HMIdimensions could be analyzed

In the last of the 30 experiments we collected the time ofthe 13 subjects to complete the take-over tasks under theseven interaction modes in Table 1 Since the tasks had beenrepeated many times the subjects were skilled and the taskcompletion time would be relatively stable After making 3Dprinting models of the side stalk and the buttons accordingto the most suitable dimensions (Table 5) the drivingsimulator was reformeden the 13 subjects were invited torepeat the seven tasks with the new HMI and this set ofcompletion time was collected To judge whether the di-mensions could effectively improve the interaction perfor-mance and efficiency of driving information processingpaired-sample t-test was used on the two groups of data eresults are shown in Table 6

It could be seen from the results that except for A2 andC2 the completion time under the two sets of dimensions inthe process of the other five tasks all had a significant dif-ference (Figure 4)

Figure 4 shows that the completion time of each task wasshorter when the dimensions with high fitness were used onthe HMI is indicated that the driverrsquos information pro-cessing was faster which helped to improve the quality oftake-over and the ability to control the car As the driverrsquosattention to the HMI might compete with the execution ofdriving tasks for cognitive resources [47] and impair thedriving performance a shorter completion time of take-overtasks was conducive to the distribution of the driverrsquospsychological resources and more beneficial to trafficefficiency

It could be found by comparing the completion time ofeach task under the new HMI that (A) for starting the self-

Table 4 Features of the 14 dimensions

Tasks anddimensions

Mean(dm)

Stddeviation

Minimum(dm)

Maximum(dm)

A1 a11 19513 00789 18086 2091a12 06055 0063 05117 06977

A2 a21 08624 00242 08224 08979a22 00784 00477 00008 01487

B1 b11 13356 00633 12528 14475b12 00315 00169 00024 00588

B2 b21 13999 00526 13060 14952b22 0322 00139 03011 03497

C1 c11 29776 01254 28032 31940c12 3272 01414 30207 34928

C2 c21 27451 00305 27010 27966c22 32714 01304 30143 34819

C3 c31 32929 00596 32019 33963c32 3244 01523 30011 34932

2500735

0074

00745

0075

00755

0076

00765

0077

00775

0078

Fittn

ess

5 10 15 200Generation

Mean fitnessBest fitness

Figure 3 Solving process of GA

8 Complexity

adaptive cruise the completion time of operating with a sidestalk was shorter (B) for the task of adjusting the distancebetween cars the completion time would be shorter whenoperating with a button located in the middle of the steeringwheel spoke (C) for the task of auto hold a starting buttonon the left side of the EPB button could bring about a shortercompletion time ese results provided a reference for thefuture automobile interior design

5 Discussion

Driving assistant system (DAS) plays an important role inthe context of human-computer codriving As reportedDAS can help the driver to complete the intense drivingtasks effectively and abate the pressure level of the driver[48] erefore it is meaningful to make a study for thecontroller dimensions in the DAS e study collectedcomfortable dimensions perceived by the drivers in a relaxedstate e aim was to detect a set of reasonable dimensions

which could make drivers at a low level of stress In ad-dition it is valuable for drivers to adapt to the different roadenvironment and realize sustainable traffic to optimize thesizes of interactive components on the in-vehicle HMI emain goals of in-vehicle technologies and cooperativeservices were to reduce traffic congestion and improve roadsafety [4]

e study attempted to analyze the optimal HMI di-mensions suitable for take-over tasks e involved HMIcomponents such as the steering wheel self-adaptive cruiselever distance adjustment button and auto hold buttonwere all direct design elements for users to see and touchBecause the design and performance of a product wereimportant factors affecting the decision-making of con-sumers [49] we calculated the information entropy andinformation gains of the three kinds of take-over tasks forclassifying usersrsquo purchase intentions e informationtransmitted by a product was multivariant complex andfuzzy so that there would be dissipation in the process oftransmission erefore the image evaluation process ofproducts could be regarded as a chaotic state e feedbackinformation could be calculated by information entropy tocarry out a comprehensive evaluation e smaller the en-tropy of an image the clearer it was From the perspective ofchaos degree the 2nd factor in Table 3 (the position of thefunction key to adjust the distance between cars) broughtmore uncertainty to the purchase decision(E(A2) 09592) which indicated that little informationwas provided by the 2nd factor for judging the usersrsquo de-cisions On the contrary the uncertainty produced by the 1stfactor (the starting mode of the self-adaptive cruise) was thesmallest (E(A1) 07842) is illustrated that using theside stalk or a button to complete the switch of the controlright had a great impact on the satisfaction and purchasedecision of users

After normalizing the information entropy of threekinds of take-over tasks their weight values were calculated

Table 5 e finally generated dimensional data of the individuals with the highest fitness

(A) e starting mode of self-adaptive cruise

(B) e position of thefunction key to adjust thedistance between cars

(C) e position of the function key of auto-hold Fitness value

a11 a12 a21 a22 b11 b12 b21 b22 c11 c12 c21 c22 c31 c32202 062 089 009 136 004 148 030 284 302 271 338 321 313 008

Table 6 e paired sample t-test results of task completion time under the two HMIs with different dimensions

Interaction modes of take-over tasks

e task completiontime under the original

dimensions (s)

e task completiontime under the evolved

dimensions (s) t-value Sig

Mean Std dev Mean Std devA1lowast 0715 0185 0537 0167 2553 0025A2 0657 0181 0637 0186 1151 0272B1lowast 0648 013 0462 0206 2797 0016B2lowastlowast 0736 0106 0571 0124 3116 0009C1lowast 0753 0295 0562 0222 2515 0027C2 0667 0226 0522 0176 1999 0069C3lowast 069 0208 0537 0225 2355 0036

A1lowast B1lowast B2lowastlowast C1lowast C3lowast

0

02

04

06

08

1

12

Time under the original dimensionsTime under the evolved dimensions

Figure 4 e completion time of the tasks with significant dif-ferences under the two HMIs with different dimensions

Complexity 9

and the fitness function was constructed In this way thedimensions with the highest fitness could be found out bymeans of GA

e background and driving experience of the subjects weredifferent and their familiarity with the take-over tasks and theHMI was variant Besides the noise of usersrsquo subjective per-ception data was always high Taking the above factors intoconsideration we collected PCD data repeatedly from the same13 subjects to form the initial population From the evolutionarytrajectory it could be found that in the 8th 10th 11th 14th 17thand 21st iteration there appeared an optimal solution

GA was always used for parameter optimization bysearching the optimal solutions while seldom used in userresearch and environment perception analysis But in studiesof user-centred design it could search the set of satisfactorysolutions driven by user expectations [46] Many pieces ofresearch had utilized GA to improve the ergonomic effi-ciency or optimize interactive component sizes in an op-erating environment [34ndash38 50 51] In this study wereported a driver perception research which investigated themost comfortable HMI sizes under different interactionmodes of take-over tasks By means of normalized geometricselection arithmetic uniform crossover and nonuniformmutation GA was applied in the research of perceivedcomfortable dimensions and a set of reasonable dimensionswas obtained e dimensional evolutionary method dis-cussed in this study embodied the concept of dynamic in-teraction design which took the mutual influence andrestrictions among different dimensions into considerationCompared with traditional methods of measuring manytimes and taking the average this method was more ad-vanced And the study enriched theoretical researches in thefield of product design and collaborative optimization

Perceived comfort was a complex problem for its dif-ficulty to be quantified objectively But it was useful in er-gonomic explorations Researchers had found that theperceptual messages from product property and usage re-quirements were competent to the analysis of user experi-ence [52] erefore it was reasonable to probe suitabledimensions by perceived comfort For the study of HMIdesign effect in a specific environment these data couldproduce important and dependable results [53] Rationaldimensions of the automobile interior HMI played an im-portant role in the driverrsquos operation quality Caruso [54]studied driversrsquo experience on the seat by an experiment andfound that interior space influenced the driversrsquo ergonomicperformance [54] For the take-over tasks in the situation ofhuman-computer codriving the dimensions also affectedwhether the driver could safely achieve the control rightswitch in the shortest possible time

In this paper paired-sample t-test was used to testwhether the sizes with the highest fitness could bring about asignificant reduction in task completion time In the resultswe found that when the interface adopted the finally evolvedsizes the completion time of each task was shorter As thecompletion time of in-vehicle interactive tasks was related tothe vehicle speed and driver errors under the context ofhuman-computer codriving [21 55] a shorter completiontime would help the driver to control the speed better

6 Limitation

e study explored the reasonable dimensions of the hu-man-machine interface in intelligent cars Neverthelessthere were still some deficiencies in the equipment used inthis study e driving simulator was easier to control andreform and helped to reduce the subjectsrsquo pressure How-ever the immersion visual interactivity and fidelity degreewere not high enough After obtaining reliable conclusionsin this study outdoor experiment with a real vehicle could beconducted in the next stage

Besides the sample size of this study was limited Whenthe sample size was larger in future studies the represen-tativeness of the data would be better and the results of GAcould be further improved

Finally usersrsquo perception of the HMI dimensions de-viated from each other In the next phase of the studyelectroencephalography (EEG) signal in the process ofperforming the driving tasks could be used By this meansthe subjectsrsquo attention and intention when interacting withthe HMI could be reflected in an objective way [56 57] as asupplement to their subjective report

7 Conclusion

For HMI design of industrial products physical concretemodelling parameters such as the product shape size andlayout were the objective properties that users could directlyperceive and the basis for the product to be expressed Basedon the dynamic measurement principle this paper proposedan optimized design method for ergonomic dimensions byGA e results indicated that the method was effective andprovided a new way for product shape innovation andhuman-machine dimensional automatic design e mainconclusions of this study were as follows

(1) Based on the information entropy theory the weightvalues of the three driving assistance functions werecalculated e results showed that the informationgain of the starting mode of the self-adaptive cruisewas highest while the information gain of the po-sition of the function key to adjust the distancebetween cars was lowest which provided a theo-retical basis for the dimensional fitness evaluationmechanism of the HMI

(2) e evolutionary process of HMI dimensionsdriven by user expectation was studied and theevolution conditions were analyzed e studyconfirmed the feasibility of normalized geometricselection arithmetic uniform crossover and non-uniform mutation in iterative researches for hu-man-machine dimensions rough parameterdebugging this evolutionary method was proved tobe effective in obtaining appropriate sizes of HMIcomponents e model could make full use of theadvantages of GA to improve the user researchperformance such as few restrictions on the opti-mization high robustness and an easily achievedglobal search

10 Complexity

(3) After finding the dimensions with the highest fitnessand applying them to the design of the HMI within-subjects experiments showed that the completiontime of five take-over tasks was significantly shorterexcept for that of Task A2 and C2is indicated thatthe information processing of drivers under the HMIwith evolved dimensions was faster which was morebeneficial to the vehicle control and take-overquality Intelligent vehicles achieved intelligent in-formation exchange with people vehicles and roadsthrough in-vehicle sensor system and informationterminals erefore the evolved dimensions couldmake the vehicle apt to respond to the environmentand have more time to fully deal with the complextraffic environment and would help to improve thetraffic efficiency therefrom

Data Availability

e data used to support the findings of this study areavailable from the corresponding author upon request

Consent

e study respected and guaranteed the subjectsrsquo right toparticipate in the study and allowed them to withdraw from thestudy unconditionally at any stage e study strictly followedthe informed consent procedure e subjects were asked tosign a consent form to inform them of the purpose processduration of the experiment matters needing attention and themethods and purpose of collecting the data so as to reducetheir anxiety and confusion In the experiment the personalsafety and health of the subjects took priority According to theusage guidelines of the driving simulator we tried to avoid anypossible injury in driving tasks e privacy of the subjects wasstrictly protected e subjects were informed of the storageuse and confidentiality measures of their personal informationand image data in advance truthfully And we would notdisclose the information to a third party or spread them on theInternet without authorization

Conflicts of Interest

e authors declare that they have no conflicts of interest

Acknowledgments

is researchwas funded by BeijingUrbanGovernance ResearchProject (grant no 20XN244) Scientific Research Program ofBeijing Education Commission (grant no KM202010009003)Yuyou Talent Support Program of North China University ofTechnology (grant no 107051360018XN012018) Tianjin YouthTalents Reserve Support Program and Chinese ErgonomicsSociety amp Kingfar Joint Research Fund for Outstanding YoungScholars (grant number CES-Kingfar-2019-001)

References

[1] S Jin J Yang E Wang and J Liu ldquoe influence of high-speed rail on ice-snow tourism in north eastern ChinardquoTourism Management vol 78 Article ID 104070 2020

[2] J Yang R Yang J Sun T Huang and Q Ge ldquoe spatialdifferentiation of the suitability of ice-snow tourist destina-tions based on a comprehensive evaluation model in ChinardquoSustainability vol 9 no 5 Article ID 774 2017

[3] W-Y Chung T-W Chong and B-G Lee ldquoMethods to detectand reduce driver stress a reviewrdquo International Journal ofAutomotive Technology vol 20 no 5 pp 1051ndash1063 2019

[4] H Farah and H N Koutsopoulos ldquoDo cooperative systemsmake driversrsquo car-following behavior saferrdquo TransportationResearch Part C Emerging Technologies vol 41 pp 61ndash722014

[5] SAE Taxonomy and Definitions for Terms Related to DrivingAutomation Systems for On-Road Motor Vehicles SAEJ3016TM Warrendale PA USA 2018

[6] J Hoegg J W Alba and D W Dahl ldquoe good the bad andthe ugly influence of aesthetics on product feature judg-mentsrdquo Journal of Consumer Psychology vol 20 no 4pp 419ndash430 2010

[7] M Kyriakidis R Happee and J C F De Winter ldquoPublicopinion on automated driving results of an internationalquestionnaire among 5000 respondentsrdquo TransportationResearch Part F Traffic Psychology and Behaviour vol 32pp 127ndash140 2015

[8] F Meng and C Spence ldquoTactile warning signals for in-vehiclesystemsrdquo Accident Analysis amp Prevention vol 75 pp 333ndash346 2015

[9] J Wan and C Wu ldquoe effects of lead time of take-overrequest and nondriving tasks on taking-over control of au-tomated vehiclesrdquo IEEE Transactions on Human MachineSystems vol 48 no 6 pp 582ndash591 2018

[10] S Petermeijer P Hornberger I Ganotis J D Winter andK Bengler ldquoe design of a vibrotactile seat for conveyingtake-over requests in automated drivingrdquo in Proceedings of the8th International Conference on Applied Human Factors andErgonomics Los Angeles CA USA July 2017

[11] T Ito A Takata and K Oosawa ldquoTime required for take-overfrom automated to manual drivingrdquo in Proceedings of the SAE2016 World Congress and Exhibition Detroit MI USA April2016

[12] K Zeeb A Buchner and M Schrauf ldquoWhat determines thetake-over time An integrated model approach of driver take-over after automated drivingrdquo Accident Analysis amp Preven-tion vol 78 pp 212ndash221 2015

[13] B Mok M Johns K J Lee et al ldquoEmergency automation offunstructured transition timing for distracted drivers of au-tomated vehiclesrdquo in Proceedings of the IEEE InternationalConference on Intelligent Transportation Systems Las PalmasSpain September 2015

[14] S Samuel A Borowsky S Zilberstein and D L FisherldquoMinimum time to situation awareness in scenarios involvingtransfer of control from an automated driving suiterdquoTransportation Research Record Journal of the TransportationResearch Board vol 2602 no 2602 pp 115ndash120 2016

[15] W Vlakveld N van Nes J de Bruin L Vissers andM van der Kroft ldquoSituation awareness increases when drivershave more time to take over the wheel in a level 3 automated

Complexity 11

car a simulator studyrdquo Transportation Research Part F TrafficPsychology and Behaviour vol 58 pp 917ndash929 2018

[16] N Merat A H Jamson F C H Lai M Daly andO M J Carsten ldquoTransition to manual driver behaviourwhen resuming control from a highly automated vehiclerdquoTransportation Research Part F Traffic Psychology and Be-haviour vol 27 no 26 pp 274ndash282 2014

[17] C Z Wu H R Wu and N C Lyu ldquoReview of control switchand safety of human-computer driving intelligent vehiclerdquoJournal of Traffic and Transportation Engineering vol 18no 6 pp 131ndash141 2018

[18] C Gold D Dambock L Lorenz and K Bengler ldquoldquoTakeoverrdquo how long does it take to get the driver back into thelooprdquo Proceedings of the Human Factors and ErgonomicsSociety Annual Meeting vol 57 no 1 pp 1938ndash1942 2013

[19] H Yang Y Zhao and Y Wang ldquoIdentifying modeling formsof instrument panel system in intelligent shared cars a studyfor perceptual preference and in-vehicle behaviorsrdquo Envi-ronmental Science and Pollution Research vol 27 no 1pp 1009ndash1023 2020

[20] A Berkovich A Lu B Levine and A V Reddy ldquoObservedcustomer seating and standing behavior and seat preferenceson board subway cars in New York cityrdquo TransportationResearch Record Journal of the Transportation ResearchBoard vol 2353 no 1 pp 33ndash46 2013

[21] R Li Y V Chen C Sha and Z Lu ldquoEffects of interface layouton the usability of in-vehicle information systems and drivingsafetyrdquo Displays vol 49 pp 124ndash132 2017

[22] H Kim S Kwon J Heo H Lee and M K Chung ldquoe effectof touch-key size on the usability of in-vehicle informationsystems and driving safety during simulated drivingrdquo AppliedErgonomics vol 45 no 3 pp 379ndash388 2014

[23] H Kim and H Song ldquoEvaluation of the safety and usability oftouch gestures in operating in-vehicle information systemswith visual occlusionrdquo Applied Ergonomics vol 45 no 3pp 789ndash798 2014

[24] M Koerber C Gold D Lechner and K Bengler ldquoe in-fluence of age on the take-over of vehicle control in highlyautomated drivingrdquo Transportation Research Part F TrafficPsychology amp Behaviour vol 39 pp 19ndash32 2016

[25] H Clark A C Mclaughlin B Williams and J Feng ldquoPer-formance in takeover and characteristics of non-driving re-lated tasks during highly automated driving in younger andolder driversrdquo Proceedings of the Human Factors amp Ergo-nomics Society Annual Meeting vol 61 no 1 pp 37ndash41 2017

[26] I Lijarcio S A Useche J Llamazares and L MontoroldquoPerceived benefits and constraints in vehicle automationdata to assess the relationship between driverrsquos features andtheir attitudes towards autonomous vehiclesrdquo Data in Briefvol 27 Article ID 104662 2019

[27] K-Z Chen and X-A Feng ldquoVirtual genes of manufacturingproducts and their reforms for product innovative designrdquoProceedings of the Institution of Mechanical Engineers Part CJournal of Mechanical Engineering Science vol 218 no 5pp 557ndash574 2004

[28] K-Z Chen X-A Feng and X-C Chen ldquoReverse deductionof virtual chromosomes of manufactured products for theirgene-engineering-based innovative designrdquo Computer-AidedDesign vol 37 no 11 pp 1191ndash1203 2005

[29] H-C Tsai and J-R Chou ldquoAutomatic design support andimage evaluation of two-coloured products using colour as-sociation and colour harmony scales and genetic algorithmrdquoComputer-Aided Design vol 39 no 9 pp 818ndash828 2007

[30] S-W Hsiao C-F Hsu and K-W Tang ldquoA consultation andsimulation system for product color planning based on in-teractive genetic algorithmsrdquo Color Research amp Applicationvol 38 no 5 pp 375ndash390 2013

[31] Y Sun W-l Wang Y-w Zhao and X-j Liu ldquoUser imageoriented interactive genetic algorithm evaluation mode inproduct developmentrdquo Computer Integrated ManufacturingSystems vol 18 no 2 pp 276ndash281 2012

[32] R Kamalian E Yeh Z Ying A M Agogino and H TakagildquoReducing human fatigue in interactive evolutionary com-putation through fuzzy systems and machine learning sys-temsrdquo in Proceedings of the 2006 IEEE InternationalConference on Fuzzy Systems Vancouver Canada July 2006

[33] Z Gu M Xi Tang and J H Frazer ldquoCapturing aestheticintention during interactive evolutionrdquo Computer-AidedDesign vol 38 no 3 pp 224ndash237 2006

[34] G Harih M Borovinsek and Z Ren Optimisation ofProductrsquos Hand-Handle Interface Material Parameters forImproved Ergonomics Springer International PublishingCham Switzerland 2015

[35] P Cheng D Chen and J Wang ldquoClustering of the bodyshape of the adult male by using principal component analysisand genetic algorithm-BP neural networkrdquo Soft Computingvol 24 no 17 Article ID 13219 2020

[36] N M B Serrano P J G Nieto A S Sanchez F S Lasherasand P R Fernandez AHybrid Algorithm for the Assessment ofthe Influence of Risk Factors in the Development of Upper LimbMusculoskeletal Disorders Springer Cham Switzerland 2018

[37] S S Sankar O-M Holman A F Gazabon and C J AcevedoldquoApplication of genetic algorithm to job scheduling underergonomic constraints in manufacturing industryrdquo Journal ofAmbient Intelligence and Humanized Computing vol 10no 5 pp 2063ndash2090 2019

[38] A-Z Atiya L Lee and K Xing ldquoDeveloping a multi-ob-jective genetic optimisation approach for an operationaldesign of a manual mixed-model assembly line with walkingworkersrdquo Journal of Intelligent Manufacturing vol 27 no 5pp 1ndash17 2014

[39] P Jedlicka and T Ryba ldquoGenetic algorithm application inimage segmentationrdquo Pattern Recognition and Image Anal-ysis vol 26 no 3 pp 497ndash501 2016

[40] L V Campenhout J Frens K Overbeeke A Standaert andH Peremans ldquoPhysical interaction in a dematerializedworldrdquo International Journal of Design vol 7 no 1 pp 1ndash182013

[41] K Nandhini and S R Balasundaram ldquoImproving readabilitythrough individualized summary extraction using interactivegenetic algorithmrdquo Applied Artificial Intelligence vol 30no 7 pp 635ndash661 2016

[42] R M Gray Entropy and Information 6eory Springer NewYork NY USA 2nd edition 2011

[43] B Anderhofstadt and S Spinler ldquoFactors affecting the pur-chasing decision and operation of alternative fuel-poweredheavy-duty trucks in Germanymdasha Delphi studyrdquo Trans-portation Research Part D Transport and Environmentvol 73 pp 87ndash107 2019

[44] X-m Wu and Y Feng ldquoReactive power optimization ofdistribution network based on improved genetic algorithmrdquoJournal of Xirsquoan Shiyou University (Natural Science Edition)vol 30 no 3 pp 95ndash99 2015

[45] Z Ling andM-j Li ldquoState space evolutionary algorithm basedon non-uniform mutation operatorrdquo Computer Technologyand Development vol 28 no 9 pp 68ndash71 2018

12 Complexity

[46] W Hu and J Zhao ldquoAutomobile styling gene evolutiondriven by usersrsquo expectation imagerdquo Journal of MechanicalEngineering vol 47 no 16 pp 176ndash181 2011

[47] J J Scott and R Gray ldquoA comparison of tactile visual andauditory warnings for rear-end collision prevention in sim-ulated drivingrdquo Human Factors 6e Journal of the HumanFactors and Ergonomics Society vol 50 no 2 pp 264ndash2752008

[48] B Reimer B Mehler and J F Coughlin ldquoReductions in self-reported stress and anticipatory heart rate with the use of asemi-automated parallel parking systemrdquo Applied Ergonom-ics vol 52 pp 120ndash127 2016

[49] B DrsquoIppolito ldquoe importance of design for firms com-petitiveness a review of the literaturerdquo Technovation vol 34no 11 pp 716ndash730 2014

[50] S Asensio-Cuesta J A Diego-Mas L Canos-Daros andC Andres-Romano ldquoA genetic algorithm for the design of jobrotation schedules considering ergonomic and competencecriteriardquo International Journal of Advanced ManufacturingTechnology vol 60 no 9ndash12 pp 1161ndash1174 2012

[51] S-C Lee H-E Tseng C-C Chang and Y-M HuangldquoApplying interactive genetic algorithms to disassembly se-quence planningrdquo International Journal of Precision Engi-neering and Manufacturing vol 21 no 4 pp 663ndash679 2020

[52] M G Shishaev V V Dikovitsky and L V Lapochkina ldquoeexperience of building cognitive user interfaces of multido-main information systems based on the mental model ofusersrdquo in Proceedings of the 6th Computer Science On-LineConference (CSOC 2017) Prague Czech Republic April 2017

[53] S Lee H Alzoubi and S Kim ldquoe effect of interior designelements and lighting layouts on prospective occupantsrsquoperceptions of amenity and efficiency in living roomsrdquo Sus-tainability vol 9 no 7 Article ID 1119 2017

[54] G Caruso ldquoMixed reality system for ergonomic assessment ofdriverrsquos seatrdquo International Journal of Virtual Reality vol 10no 2 pp 69ndash79 2011

[55] M L Reyes and J D Lee ldquoEffects of cognitive load presenceand duration on driver eye movements and event detectionperformancerdquo Transportation Research Part F Traffic Psy-chology and Behaviour vol 11 no 6 pp 391ndash402 2008

[56] D Cisler P M Greenwood D M Roberts R McKendrickand C L Baldwin ldquoComparing the relative strengths of EEGand low-cost physiological devices in modeling attentionallocation in semiautonomous vehiclesrdquo Frontiers in HumanNeuroscience vol 13 Article ID 109 2019

[57] I-H Kim J-W Kim S Haufe and S-W Lee ldquoDetection ofbraking intention in diverse situations during simulateddriving based on EEG feature combinationrdquo Journal of NeuralEngineering vol 12 no 1 Article ID 016001 2014

Complexity 13

323 Genetic Operation and Evolutionary Process In thissystem the premise to debug the genetic operation andmakejudgment was due to the irrationality and uncertainty of theshape and spatial dimension design of automobile con-trollers and panels the dimension evolution driven by usersrsquoexpectation was not only to search for the optimal solutionbut also to obtain the satisfactory solution under limitedresources erefore the model was different in codingselection crossover and mutation from traditional GAwhich was to solve problems represented by explicitfunctions

(1) Coding Generally there are two kinds of codingmethods for optimization problems which are real numbercoding and binary coding And both have advantages anddisadvantages Mapping errors exist in traditional binarycoding when continuous functions are discretizedWhen theindividual coding string is short the accuracy requirementsmay not be met While when the string is long although thecoding accuracy can be improved the search space of GAwill expand dramatically [44] erefore we used realnumber coding in this study

In the process of evolution a population Pop(t) wasformed Each chromosome Xt

i in the population representeddata from one of the 30 experiments e genes in thechromosome were composed of PCD reported by thesubjects after completing the whole experiment processwhich could be expressed as follows

Pop(t) Xt1 X

t2 X

ti X

tM1113966 1113967

Xti θt

i1 θti2 θt

ib θtiN1113960 1113961

(3)

where t represented the generation number of evolutions θrepresented gene coding and M represented the populationsize

(2) Initialized Population and Selection e populationneeded to be initialized before running GA According to theresearch demand PCD data from the 30 experiments weretaken as the primary population Each time data of the 14PCD indexes were collected to form the population matrix

e selection operation was the process of selectingindividuals from the previous generation of the populationto form the next generation e purpose of selection was toobtain fine individuals based on the fitness values so thatthey would have the opportunity to reproduce as parents forthe next generation Individuals with high fitness were morelikely to be inherited while the ones with low fitness wereless likely After debugging and comparison we chose themethod of normalized geometric select (NGS) to performthe selection operation NGS was mainly to sort the fitnessvalues and the better individuals would be maintained asparents is was more suitable for the selection of productdesign schemes and also helpful to prevent better individualsfrom being damaged

(3) Crossover and Mutation By crossover operation anew generation of individuals could be obtained whichcombined the characteristics of their parents and embodiedthe idea of information exchange In the case of real number

coding the arithmetic uniform crossover (AUC) betweenindividuals was generally utilized [45] AUC was a linearcombination of two individuals to produce two new onesemethod of crossover and getting the next generation wasas follows

c1 p1lowast a + p2lowast (1 minus a)

c2 p1lowast (1 minus a) + p2lowast a(4)

where p1 and p2 were the parents and a was a random mixamount

In this study the nonuniform mutation (NM) was usedto perform the mutation operation Let an individual beX X1X2 Xk Xl If Xk was the variation point andits value range was [Uk

min Ukmax] a new individual

X X1X2 Xkprime Xl could be obtained after non-

uniformly mutating at this point And the new gene valuewas given by

Xkprime

Xk + Δ t Ukmax minus Xk1113872 1113873 if random(0 1) 0

Xk minus Δ t Xk minus Ukmin1113872 1113873 if random(0 1) 1

⎧⎨

⎩

(5)

where Δ(t y) was a random number conforming to thenonuniform distribution in the range [0 y] y representedUk

max minus Xk and Xk minus Ukmin It was required that with the

increase of evolutionary generation number t the proba-bility of Δ(t y) approaching to 0 also raised gradually

In this study Δ(t y) was defined as follows

Δ(t y) y 1 minus r(1minus tT)b

1113872 1113873 (6)

where r was a random number conforming to the non-uniform distribution in the range [0 1] which was theselection pressure T was the maximum number of evolu-tionary generations in this study T 25 In the 25 iterationsthe better individuals with high fitness could be found out toexplore the evolutionary mechanism of the appropriate HMIdimensions of the DAFs b was a parameter to adjust thevariable step size which was a system parameter It deter-mined the dependence degree of the random number dis-turbance on t and the range was usually 2sim5 Afterdebugging the model and making comparisons repeatedlythis parameter was set as b 4 in this study

4 Results

41 Information Entropy and the Weights of the DAFs

4116e Expectation Information for the Classification of theHMI of Take-Over Tasks According to different combina-tions of the seven interaction modes in Table 1 a total of2lowast 2lowast 312 HMI schemes could be generated By means ofDelphi method purchase decisions for the 12 schemes wereinvestigated After four rounds of expert investigations thepurchase decisions of the eight professional users reached anagreement e classification results are shown in Table 2

According to the purchase intentions of the professionalusers 12 HMI schemes could be divided into two categories

6 Complexity

U1 1 (buy) and U2 2 (not buy) By summarizing andanalyzing the data the probabilities of the two categorieswere as follows P (U1) (512) and P (U2) (712) Basedon the formula the total entropy was as follows

I (U) minus512log2

512

minus712log2

712

09799 (7)

412 6e Conditional Entropy and Information Gain of EachInfluencing Factor In this study there were three factorsthat determined the purchase intentions of users A1 thestarting mode of the self-adaptive cruise A2 the position ofthe function key to adjust the distance between cars and A3the position of the function key of auto hold e proba-bilities of the factors and each condition are shown inTable 3

e information entropy values of the three factors wereas follows

E A1( 1113857 612

I(1 5) +612

I(4 2) 07842

E A2( 1113857 612

I(3 3) +612

I(2 4) 09592

E A3( 1113857 412

I(1 3) +412

I(1 3) +412

I(3 1) 08113

(8)

erefore the information gains of the factors were asfollows

Gain A1( 1113857 I(U) minus E A1( 1113857 01957

Gain A2( 1113857 I(U) minus E A2( 1113857 00207

Gain A3( 1113857 I(U) minus E A3( 1113857 01686

(9)

According to the information gains the starting mode ofthe self-adaptive cruise was the most important for the userrsquosperceived comfort and purchase intension while the posi-tion of the function key to adjust the distance between carswas not so significant By normalizing the information gainsof each influencing factor the results were taken as theweights of the factors which were

β1 Gain A1( 1113857

Gain A1( 1113857 + Gain A2( 1113857 + Gain A3( 1113857 05083

β2 Gain A2( 1113857

Gain A1( 1113857 + Gain A2( 1113857 + Gain A3( 1113857 00538

β3 Gain A3( 1113857

Gain A1( 1113857 + Gain A2( 1113857 + Gain A3( 1113857 04379

(10)

After taking these values into formula (1) the artificialfitness function could be constructed as follows

F 1

05083lowastQ1 + 00538lowastQ2 + 04379lowastQ3( 1113857 (11)

42 Results of GA

421 Characteristics of the Population e population sizeof this study was Pop 30 For the convenience of recordingthe unit of each dimension was decimeter (dm)e featuresof the 14 dimensions are described in Table 4

422 Solving Process of GA To match the userrsquos expecta-tion the seed of each initial population should be able to

Table 2 Classification of purchase intentions to interaction modes

e starting mode of self-adaptive cruise

e position of the function key to adjust thedistance between cars

e position of the function key ofauto-hold

Purchasedecision

Pulling the side stalk On the middle of the left spoke Under the EPB button 2Pressing the button On the middle of the left spoke Under the EPB button 2Pulling the side stalk On the bottom edge of the left spoke Under the EPB button 2Pressing the button On the bottom edge of the left spoke Under the EPB button 1Pulling the side stalk On the middle of the left spoke On the left of the EPB button 2Pressing the button On the middle of the left spoke On the left of the EPB button 1Pulling the side stalk On the bottom edge of the left spoke On the left of the EPB button 2Pressing the button On the bottom edge of the left spoke On the left of the EPB button 2Pulling the side stalk On the middle of the left spoke On the right of the EPB button 1Pressing the button On the middle of the left spoke On the right of the EPB button 1Pulling the side stalk On the bottom edge of the left spoke On the right of the EPB button 2Pressing the button On the bottom edge of the left spoke On the right of the EPB button 1

Table 3 e probabilities of the three factors and each condition

P (Vi) P (U1 | Vi) P (U2 | Vi)

A1Side stalk 612 16 56Button 612 46 26A2Middle 612 36 36Bottom edge 612 26 46A3Under the EPB button 412 14 34On the left of the EPB button 412 14 34On the right of the EPB button 412 34 14P (Vi) represents the probability of each condition of the influencingfactors P(U1 | Vi) and P(U2 | Vi) respectively indicate the conditionalprobabilities of the two categories

Complexity 7

cover part of the information of the PCD data And it shouldbe capable of satisfying the geometric description rules ofproduct evolution in terms of space and size [46] e initialpopulation of this paper came from repeated experiments inmany days and the perceptual experience and spatial at-tributes reflected by the data were stable e optimal so-lution of the model was obtained after 25 iterations Figure 3shows the evolutionary process of GA

In the figure the red line represents the best fitness valueof each generation And the blue dotted line describes themean fitness of each generation It could be seen from thefigure that the optimal value in the model increased with thenumber of iterations and converged into a straight line afterreaching the maximum value Until the optimal value of themodel remained unchanged the global optimal solutioncould be obtained

e genetic evolution stopped after 25 iterations It couldbe seen from the figure that although the deviation betweenthe optimal fitness values and the mean ones increased afterthe 9th iteration it continued to evolve which met theexpectation of evolution After the 21st iteration the devi-ation was close to 0

In the 25 generations of evolutionary iteration based onthe PCD data eight individuals with the highest fitness wereobtained which were the HMI dimension schemes that metthe usersrsquo expectation best in a certain generation of thepopulation It would be rational to make designs of HMIssuch as steering wheels and gear lever panels in the laterstage according to these fine values is could not onlyinherit the modelling genes of the HMIs but was alsocoupled to the user expectation to a certain extent e fi-nally generated dimensional data of the individuals with thehighest fitness are shown in Table 5

43 Significance of the Difference between the Original Di-mensions and the Evolved Ones To further validate whetherthe individuals with the highest fitness could bring about a

better take-over performance it was reasonable to collect theneeded time to complete driving assistant tasks under theHMI with the original sizes and the one with the evolvedsizes And the differences between the two sets of HMIdimensions could be analyzed

In the last of the 30 experiments we collected the time ofthe 13 subjects to complete the take-over tasks under theseven interaction modes in Table 1 Since the tasks had beenrepeated many times the subjects were skilled and the taskcompletion time would be relatively stable After making 3Dprinting models of the side stalk and the buttons accordingto the most suitable dimensions (Table 5) the drivingsimulator was reformeden the 13 subjects were invited torepeat the seven tasks with the new HMI and this set ofcompletion time was collected To judge whether the di-mensions could effectively improve the interaction perfor-mance and efficiency of driving information processingpaired-sample t-test was used on the two groups of data eresults are shown in Table 6

It could be seen from the results that except for A2 andC2 the completion time under the two sets of dimensions inthe process of the other five tasks all had a significant dif-ference (Figure 4)

Figure 4 shows that the completion time of each task wasshorter when the dimensions with high fitness were used onthe HMI is indicated that the driverrsquos information pro-cessing was faster which helped to improve the quality oftake-over and the ability to control the car As the driverrsquosattention to the HMI might compete with the execution ofdriving tasks for cognitive resources [47] and impair thedriving performance a shorter completion time of take-overtasks was conducive to the distribution of the driverrsquospsychological resources and more beneficial to trafficefficiency

It could be found by comparing the completion time ofeach task under the new HMI that (A) for starting the self-

Table 4 Features of the 14 dimensions

Tasks anddimensions

Mean(dm)

Stddeviation

Minimum(dm)

Maximum(dm)

A1 a11 19513 00789 18086 2091a12 06055 0063 05117 06977

A2 a21 08624 00242 08224 08979a22 00784 00477 00008 01487

B1 b11 13356 00633 12528 14475b12 00315 00169 00024 00588

B2 b21 13999 00526 13060 14952b22 0322 00139 03011 03497

C1 c11 29776 01254 28032 31940c12 3272 01414 30207 34928

C2 c21 27451 00305 27010 27966c22 32714 01304 30143 34819

C3 c31 32929 00596 32019 33963c32 3244 01523 30011 34932

2500735

0074

00745

0075

00755

0076

00765

0077

00775

0078

Fittn

ess

5 10 15 200Generation

Mean fitnessBest fitness

Figure 3 Solving process of GA

8 Complexity

adaptive cruise the completion time of operating with a sidestalk was shorter (B) for the task of adjusting the distancebetween cars the completion time would be shorter whenoperating with a button located in the middle of the steeringwheel spoke (C) for the task of auto hold a starting buttonon the left side of the EPB button could bring about a shortercompletion time ese results provided a reference for thefuture automobile interior design

5 Discussion

Driving assistant system (DAS) plays an important role inthe context of human-computer codriving As reportedDAS can help the driver to complete the intense drivingtasks effectively and abate the pressure level of the driver[48] erefore it is meaningful to make a study for thecontroller dimensions in the DAS e study collectedcomfortable dimensions perceived by the drivers in a relaxedstate e aim was to detect a set of reasonable dimensions

which could make drivers at a low level of stress In ad-dition it is valuable for drivers to adapt to the different roadenvironment and realize sustainable traffic to optimize thesizes of interactive components on the in-vehicle HMI emain goals of in-vehicle technologies and cooperativeservices were to reduce traffic congestion and improve roadsafety [4]

e study attempted to analyze the optimal HMI di-mensions suitable for take-over tasks e involved HMIcomponents such as the steering wheel self-adaptive cruiselever distance adjustment button and auto hold buttonwere all direct design elements for users to see and touchBecause the design and performance of a product wereimportant factors affecting the decision-making of con-sumers [49] we calculated the information entropy andinformation gains of the three kinds of take-over tasks forclassifying usersrsquo purchase intentions e informationtransmitted by a product was multivariant complex andfuzzy so that there would be dissipation in the process oftransmission erefore the image evaluation process ofproducts could be regarded as a chaotic state e feedbackinformation could be calculated by information entropy tocarry out a comprehensive evaluation e smaller the en-tropy of an image the clearer it was From the perspective ofchaos degree the 2nd factor in Table 3 (the position of thefunction key to adjust the distance between cars) broughtmore uncertainty to the purchase decision(E(A2) 09592) which indicated that little informationwas provided by the 2nd factor for judging the usersrsquo de-cisions On the contrary the uncertainty produced by the 1stfactor (the starting mode of the self-adaptive cruise) was thesmallest (E(A1) 07842) is illustrated that using theside stalk or a button to complete the switch of the controlright had a great impact on the satisfaction and purchasedecision of users

After normalizing the information entropy of threekinds of take-over tasks their weight values were calculated

Table 5 e finally generated dimensional data of the individuals with the highest fitness

(A) e starting mode of self-adaptive cruise

(B) e position of thefunction key to adjust thedistance between cars

(C) e position of the function key of auto-hold Fitness value

a11 a12 a21 a22 b11 b12 b21 b22 c11 c12 c21 c22 c31 c32202 062 089 009 136 004 148 030 284 302 271 338 321 313 008

Table 6 e paired sample t-test results of task completion time under the two HMIs with different dimensions

Interaction modes of take-over tasks

e task completiontime under the original

dimensions (s)

e task completiontime under the evolved

dimensions (s) t-value Sig

Mean Std dev Mean Std devA1lowast 0715 0185 0537 0167 2553 0025A2 0657 0181 0637 0186 1151 0272B1lowast 0648 013 0462 0206 2797 0016B2lowastlowast 0736 0106 0571 0124 3116 0009C1lowast 0753 0295 0562 0222 2515 0027C2 0667 0226 0522 0176 1999 0069C3lowast 069 0208 0537 0225 2355 0036

A1lowast B1lowast B2lowastlowast C1lowast C3lowast

0

02

04

06

08

1

12

Time under the original dimensionsTime under the evolved dimensions

Figure 4 e completion time of the tasks with significant dif-ferences under the two HMIs with different dimensions

Complexity 9

and the fitness function was constructed In this way thedimensions with the highest fitness could be found out bymeans of GA

e background and driving experience of the subjects weredifferent and their familiarity with the take-over tasks and theHMI was variant Besides the noise of usersrsquo subjective per-ception data was always high Taking the above factors intoconsideration we collected PCD data repeatedly from the same13 subjects to form the initial population From the evolutionarytrajectory it could be found that in the 8th 10th 11th 14th 17thand 21st iteration there appeared an optimal solution

GA was always used for parameter optimization bysearching the optimal solutions while seldom used in userresearch and environment perception analysis But in studiesof user-centred design it could search the set of satisfactorysolutions driven by user expectations [46] Many pieces ofresearch had utilized GA to improve the ergonomic effi-ciency or optimize interactive component sizes in an op-erating environment [34ndash38 50 51] In this study wereported a driver perception research which investigated themost comfortable HMI sizes under different interactionmodes of take-over tasks By means of normalized geometricselection arithmetic uniform crossover and nonuniformmutation GA was applied in the research of perceivedcomfortable dimensions and a set of reasonable dimensionswas obtained e dimensional evolutionary method dis-cussed in this study embodied the concept of dynamic in-teraction design which took the mutual influence andrestrictions among different dimensions into considerationCompared with traditional methods of measuring manytimes and taking the average this method was more ad-vanced And the study enriched theoretical researches in thefield of product design and collaborative optimization

Perceived comfort was a complex problem for its dif-ficulty to be quantified objectively But it was useful in er-gonomic explorations Researchers had found that theperceptual messages from product property and usage re-quirements were competent to the analysis of user experi-ence [52] erefore it was reasonable to probe suitabledimensions by perceived comfort For the study of HMIdesign effect in a specific environment these data couldproduce important and dependable results [53] Rationaldimensions of the automobile interior HMI played an im-portant role in the driverrsquos operation quality Caruso [54]studied driversrsquo experience on the seat by an experiment andfound that interior space influenced the driversrsquo ergonomicperformance [54] For the take-over tasks in the situation ofhuman-computer codriving the dimensions also affectedwhether the driver could safely achieve the control rightswitch in the shortest possible time

In this paper paired-sample t-test was used to testwhether the sizes with the highest fitness could bring about asignificant reduction in task completion time In the resultswe found that when the interface adopted the finally evolvedsizes the completion time of each task was shorter As thecompletion time of in-vehicle interactive tasks was related tothe vehicle speed and driver errors under the context ofhuman-computer codriving [21 55] a shorter completiontime would help the driver to control the speed better

6 Limitation

e study explored the reasonable dimensions of the hu-man-machine interface in intelligent cars Neverthelessthere were still some deficiencies in the equipment used inthis study e driving simulator was easier to control andreform and helped to reduce the subjectsrsquo pressure How-ever the immersion visual interactivity and fidelity degreewere not high enough After obtaining reliable conclusionsin this study outdoor experiment with a real vehicle could beconducted in the next stage

Besides the sample size of this study was limited Whenthe sample size was larger in future studies the represen-tativeness of the data would be better and the results of GAcould be further improved

Finally usersrsquo perception of the HMI dimensions de-viated from each other In the next phase of the studyelectroencephalography (EEG) signal in the process ofperforming the driving tasks could be used By this meansthe subjectsrsquo attention and intention when interacting withthe HMI could be reflected in an objective way [56 57] as asupplement to their subjective report

7 Conclusion

For HMI design of industrial products physical concretemodelling parameters such as the product shape size andlayout were the objective properties that users could directlyperceive and the basis for the product to be expressed Basedon the dynamic measurement principle this paper proposedan optimized design method for ergonomic dimensions byGA e results indicated that the method was effective andprovided a new way for product shape innovation andhuman-machine dimensional automatic design e mainconclusions of this study were as follows

(1) Based on the information entropy theory the weightvalues of the three driving assistance functions werecalculated e results showed that the informationgain of the starting mode of the self-adaptive cruisewas highest while the information gain of the po-sition of the function key to adjust the distancebetween cars was lowest which provided a theo-retical basis for the dimensional fitness evaluationmechanism of the HMI

(2) e evolutionary process of HMI dimensionsdriven by user expectation was studied and theevolution conditions were analyzed e studyconfirmed the feasibility of normalized geometricselection arithmetic uniform crossover and non-uniform mutation in iterative researches for hu-man-machine dimensions rough parameterdebugging this evolutionary method was proved tobe effective in obtaining appropriate sizes of HMIcomponents e model could make full use of theadvantages of GA to improve the user researchperformance such as few restrictions on the opti-mization high robustness and an easily achievedglobal search

10 Complexity

(3) After finding the dimensions with the highest fitnessand applying them to the design of the HMI within-subjects experiments showed that the completiontime of five take-over tasks was significantly shorterexcept for that of Task A2 and C2is indicated thatthe information processing of drivers under the HMIwith evolved dimensions was faster which was morebeneficial to the vehicle control and take-overquality Intelligent vehicles achieved intelligent in-formation exchange with people vehicles and roadsthrough in-vehicle sensor system and informationterminals erefore the evolved dimensions couldmake the vehicle apt to respond to the environmentand have more time to fully deal with the complextraffic environment and would help to improve thetraffic efficiency therefrom

Data Availability

e data used to support the findings of this study areavailable from the corresponding author upon request

Consent

e study respected and guaranteed the subjectsrsquo right toparticipate in the study and allowed them to withdraw from thestudy unconditionally at any stage e study strictly followedthe informed consent procedure e subjects were asked tosign a consent form to inform them of the purpose processduration of the experiment matters needing attention and themethods and purpose of collecting the data so as to reducetheir anxiety and confusion In the experiment the personalsafety and health of the subjects took priority According to theusage guidelines of the driving simulator we tried to avoid anypossible injury in driving tasks e privacy of the subjects wasstrictly protected e subjects were informed of the storageuse and confidentiality measures of their personal informationand image data in advance truthfully And we would notdisclose the information to a third party or spread them on theInternet without authorization

Conflicts of Interest

e authors declare that they have no conflicts of interest

Acknowledgments

is researchwas funded by BeijingUrbanGovernance ResearchProject (grant no 20XN244) Scientific Research Program ofBeijing Education Commission (grant no KM202010009003)Yuyou Talent Support Program of North China University ofTechnology (grant no 107051360018XN012018) Tianjin YouthTalents Reserve Support Program and Chinese ErgonomicsSociety amp Kingfar Joint Research Fund for Outstanding YoungScholars (grant number CES-Kingfar-2019-001)

References

[1] S Jin J Yang E Wang and J Liu ldquoe influence of high-speed rail on ice-snow tourism in north eastern ChinardquoTourism Management vol 78 Article ID 104070 2020

[2] J Yang R Yang J Sun T Huang and Q Ge ldquoe spatialdifferentiation of the suitability of ice-snow tourist destina-tions based on a comprehensive evaluation model in ChinardquoSustainability vol 9 no 5 Article ID 774 2017

[3] W-Y Chung T-W Chong and B-G Lee ldquoMethods to detectand reduce driver stress a reviewrdquo International Journal ofAutomotive Technology vol 20 no 5 pp 1051ndash1063 2019

[4] H Farah and H N Koutsopoulos ldquoDo cooperative systemsmake driversrsquo car-following behavior saferrdquo TransportationResearch Part C Emerging Technologies vol 41 pp 61ndash722014

[5] SAE Taxonomy and Definitions for Terms Related to DrivingAutomation Systems for On-Road Motor Vehicles SAEJ3016TM Warrendale PA USA 2018

[6] J Hoegg J W Alba and D W Dahl ldquoe good the bad andthe ugly influence of aesthetics on product feature judg-mentsrdquo Journal of Consumer Psychology vol 20 no 4pp 419ndash430 2010

[7] M Kyriakidis R Happee and J C F De Winter ldquoPublicopinion on automated driving results of an internationalquestionnaire among 5000 respondentsrdquo TransportationResearch Part F Traffic Psychology and Behaviour vol 32pp 127ndash140 2015

[8] F Meng and C Spence ldquoTactile warning signals for in-vehiclesystemsrdquo Accident Analysis amp Prevention vol 75 pp 333ndash346 2015

[9] J Wan and C Wu ldquoe effects of lead time of take-overrequest and nondriving tasks on taking-over control of au-tomated vehiclesrdquo IEEE Transactions on Human MachineSystems vol 48 no 6 pp 582ndash591 2018

[10] S Petermeijer P Hornberger I Ganotis J D Winter andK Bengler ldquoe design of a vibrotactile seat for conveyingtake-over requests in automated drivingrdquo in Proceedings of the8th International Conference on Applied Human Factors andErgonomics Los Angeles CA USA July 2017

[11] T Ito A Takata and K Oosawa ldquoTime required for take-overfrom automated to manual drivingrdquo in Proceedings of the SAE2016 World Congress and Exhibition Detroit MI USA April2016

[12] K Zeeb A Buchner and M Schrauf ldquoWhat determines thetake-over time An integrated model approach of driver take-over after automated drivingrdquo Accident Analysis amp Preven-tion vol 78 pp 212ndash221 2015

[13] B Mok M Johns K J Lee et al ldquoEmergency automation offunstructured transition timing for distracted drivers of au-tomated vehiclesrdquo in Proceedings of the IEEE InternationalConference on Intelligent Transportation Systems Las PalmasSpain September 2015

[14] S Samuel A Borowsky S Zilberstein and D L FisherldquoMinimum time to situation awareness in scenarios involvingtransfer of control from an automated driving suiterdquoTransportation Research Record Journal of the TransportationResearch Board vol 2602 no 2602 pp 115ndash120 2016

[15] W Vlakveld N van Nes J de Bruin L Vissers andM van der Kroft ldquoSituation awareness increases when drivershave more time to take over the wheel in a level 3 automated

Complexity 11

car a simulator studyrdquo Transportation Research Part F TrafficPsychology and Behaviour vol 58 pp 917ndash929 2018

[16] N Merat A H Jamson F C H Lai M Daly andO M J Carsten ldquoTransition to manual driver behaviourwhen resuming control from a highly automated vehiclerdquoTransportation Research Part F Traffic Psychology and Be-haviour vol 27 no 26 pp 274ndash282 2014

[17] C Z Wu H R Wu and N C Lyu ldquoReview of control switchand safety of human-computer driving intelligent vehiclerdquoJournal of Traffic and Transportation Engineering vol 18no 6 pp 131ndash141 2018

[18] C Gold D Dambock L Lorenz and K Bengler ldquoldquoTakeoverrdquo how long does it take to get the driver back into thelooprdquo Proceedings of the Human Factors and ErgonomicsSociety Annual Meeting vol 57 no 1 pp 1938ndash1942 2013

[19] H Yang Y Zhao and Y Wang ldquoIdentifying modeling formsof instrument panel system in intelligent shared cars a studyfor perceptual preference and in-vehicle behaviorsrdquo Envi-ronmental Science and Pollution Research vol 27 no 1pp 1009ndash1023 2020

[20] A Berkovich A Lu B Levine and A V Reddy ldquoObservedcustomer seating and standing behavior and seat preferenceson board subway cars in New York cityrdquo TransportationResearch Record Journal of the Transportation ResearchBoard vol 2353 no 1 pp 33ndash46 2013

[21] R Li Y V Chen C Sha and Z Lu ldquoEffects of interface layouton the usability of in-vehicle information systems and drivingsafetyrdquo Displays vol 49 pp 124ndash132 2017

[22] H Kim S Kwon J Heo H Lee and M K Chung ldquoe effectof touch-key size on the usability of in-vehicle informationsystems and driving safety during simulated drivingrdquo AppliedErgonomics vol 45 no 3 pp 379ndash388 2014

[23] H Kim and H Song ldquoEvaluation of the safety and usability oftouch gestures in operating in-vehicle information systemswith visual occlusionrdquo Applied Ergonomics vol 45 no 3pp 789ndash798 2014

[24] M Koerber C Gold D Lechner and K Bengler ldquoe in-fluence of age on the take-over of vehicle control in highlyautomated drivingrdquo Transportation Research Part F TrafficPsychology amp Behaviour vol 39 pp 19ndash32 2016

[25] H Clark A C Mclaughlin B Williams and J Feng ldquoPer-formance in takeover and characteristics of non-driving re-lated tasks during highly automated driving in younger andolder driversrdquo Proceedings of the Human Factors amp Ergo-nomics Society Annual Meeting vol 61 no 1 pp 37ndash41 2017

[26] I Lijarcio S A Useche J Llamazares and L MontoroldquoPerceived benefits and constraints in vehicle automationdata to assess the relationship between driverrsquos features andtheir attitudes towards autonomous vehiclesrdquo Data in Briefvol 27 Article ID 104662 2019

[27] K-Z Chen and X-A Feng ldquoVirtual genes of manufacturingproducts and their reforms for product innovative designrdquoProceedings of the Institution of Mechanical Engineers Part CJournal of Mechanical Engineering Science vol 218 no 5pp 557ndash574 2004

[28] K-Z Chen X-A Feng and X-C Chen ldquoReverse deductionof virtual chromosomes of manufactured products for theirgene-engineering-based innovative designrdquo Computer-AidedDesign vol 37 no 11 pp 1191ndash1203 2005

[29] H-C Tsai and J-R Chou ldquoAutomatic design support andimage evaluation of two-coloured products using colour as-sociation and colour harmony scales and genetic algorithmrdquoComputer-Aided Design vol 39 no 9 pp 818ndash828 2007

[30] S-W Hsiao C-F Hsu and K-W Tang ldquoA consultation andsimulation system for product color planning based on in-teractive genetic algorithmsrdquo Color Research amp Applicationvol 38 no 5 pp 375ndash390 2013

[31] Y Sun W-l Wang Y-w Zhao and X-j Liu ldquoUser imageoriented interactive genetic algorithm evaluation mode inproduct developmentrdquo Computer Integrated ManufacturingSystems vol 18 no 2 pp 276ndash281 2012

[32] R Kamalian E Yeh Z Ying A M Agogino and H TakagildquoReducing human fatigue in interactive evolutionary com-putation through fuzzy systems and machine learning sys-temsrdquo in Proceedings of the 2006 IEEE InternationalConference on Fuzzy Systems Vancouver Canada July 2006

[33] Z Gu M Xi Tang and J H Frazer ldquoCapturing aestheticintention during interactive evolutionrdquo Computer-AidedDesign vol 38 no 3 pp 224ndash237 2006

[34] G Harih M Borovinsek and Z Ren Optimisation ofProductrsquos Hand-Handle Interface Material Parameters forImproved Ergonomics Springer International PublishingCham Switzerland 2015

[35] P Cheng D Chen and J Wang ldquoClustering of the bodyshape of the adult male by using principal component analysisand genetic algorithm-BP neural networkrdquo Soft Computingvol 24 no 17 Article ID 13219 2020

[36] N M B Serrano P J G Nieto A S Sanchez F S Lasherasand P R Fernandez AHybrid Algorithm for the Assessment ofthe Influence of Risk Factors in the Development of Upper LimbMusculoskeletal Disorders Springer Cham Switzerland 2018

[37] S S Sankar O-M Holman A F Gazabon and C J AcevedoldquoApplication of genetic algorithm to job scheduling underergonomic constraints in manufacturing industryrdquo Journal ofAmbient Intelligence and Humanized Computing vol 10no 5 pp 2063ndash2090 2019

[38] A-Z Atiya L Lee and K Xing ldquoDeveloping a multi-ob-jective genetic optimisation approach for an operationaldesign of a manual mixed-model assembly line with walkingworkersrdquo Journal of Intelligent Manufacturing vol 27 no 5pp 1ndash17 2014

[39] P Jedlicka and T Ryba ldquoGenetic algorithm application inimage segmentationrdquo Pattern Recognition and Image Anal-ysis vol 26 no 3 pp 497ndash501 2016

[40] L V Campenhout J Frens K Overbeeke A Standaert andH Peremans ldquoPhysical interaction in a dematerializedworldrdquo International Journal of Design vol 7 no 1 pp 1ndash182013

[41] K Nandhini and S R Balasundaram ldquoImproving readabilitythrough individualized summary extraction using interactivegenetic algorithmrdquo Applied Artificial Intelligence vol 30no 7 pp 635ndash661 2016

[42] R M Gray Entropy and Information 6eory Springer NewYork NY USA 2nd edition 2011

[43] B Anderhofstadt and S Spinler ldquoFactors affecting the pur-chasing decision and operation of alternative fuel-poweredheavy-duty trucks in Germanymdasha Delphi studyrdquo Trans-portation Research Part D Transport and Environmentvol 73 pp 87ndash107 2019

[44] X-m Wu and Y Feng ldquoReactive power optimization ofdistribution network based on improved genetic algorithmrdquoJournal of Xirsquoan Shiyou University (Natural Science Edition)vol 30 no 3 pp 95ndash99 2015

[45] Z Ling andM-j Li ldquoState space evolutionary algorithm basedon non-uniform mutation operatorrdquo Computer Technologyand Development vol 28 no 9 pp 68ndash71 2018

12 Complexity

[46] W Hu and J Zhao ldquoAutomobile styling gene evolutiondriven by usersrsquo expectation imagerdquo Journal of MechanicalEngineering vol 47 no 16 pp 176ndash181 2011

[47] J J Scott and R Gray ldquoA comparison of tactile visual andauditory warnings for rear-end collision prevention in sim-ulated drivingrdquo Human Factors 6e Journal of the HumanFactors and Ergonomics Society vol 50 no 2 pp 264ndash2752008

[48] B Reimer B Mehler and J F Coughlin ldquoReductions in self-reported stress and anticipatory heart rate with the use of asemi-automated parallel parking systemrdquo Applied Ergonom-ics vol 52 pp 120ndash127 2016

[49] B DrsquoIppolito ldquoe importance of design for firms com-petitiveness a review of the literaturerdquo Technovation vol 34no 11 pp 716ndash730 2014

[50] S Asensio-Cuesta J A Diego-Mas L Canos-Daros andC Andres-Romano ldquoA genetic algorithm for the design of jobrotation schedules considering ergonomic and competencecriteriardquo International Journal of Advanced ManufacturingTechnology vol 60 no 9ndash12 pp 1161ndash1174 2012

[51] S-C Lee H-E Tseng C-C Chang and Y-M HuangldquoApplying interactive genetic algorithms to disassembly se-quence planningrdquo International Journal of Precision Engi-neering and Manufacturing vol 21 no 4 pp 663ndash679 2020

[52] M G Shishaev V V Dikovitsky and L V Lapochkina ldquoeexperience of building cognitive user interfaces of multido-main information systems based on the mental model ofusersrdquo in Proceedings of the 6th Computer Science On-LineConference (CSOC 2017) Prague Czech Republic April 2017

[53] S Lee H Alzoubi and S Kim ldquoe effect of interior designelements and lighting layouts on prospective occupantsrsquoperceptions of amenity and efficiency in living roomsrdquo Sus-tainability vol 9 no 7 Article ID 1119 2017

[54] G Caruso ldquoMixed reality system for ergonomic assessment ofdriverrsquos seatrdquo International Journal of Virtual Reality vol 10no 2 pp 69ndash79 2011

[55] M L Reyes and J D Lee ldquoEffects of cognitive load presenceand duration on driver eye movements and event detectionperformancerdquo Transportation Research Part F Traffic Psy-chology and Behaviour vol 11 no 6 pp 391ndash402 2008

[56] D Cisler P M Greenwood D M Roberts R McKendrickand C L Baldwin ldquoComparing the relative strengths of EEGand low-cost physiological devices in modeling attentionallocation in semiautonomous vehiclesrdquo Frontiers in HumanNeuroscience vol 13 Article ID 109 2019

[57] I-H Kim J-W Kim S Haufe and S-W Lee ldquoDetection ofbraking intention in diverse situations during simulateddriving based on EEG feature combinationrdquo Journal of NeuralEngineering vol 12 no 1 Article ID 016001 2014

Complexity 13

U1 1 (buy) and U2 2 (not buy) By summarizing andanalyzing the data the probabilities of the two categorieswere as follows P (U1) (512) and P (U2) (712) Basedon the formula the total entropy was as follows

I (U) minus512log2

512

minus712log2

712

09799 (7)

412 6e Conditional Entropy and Information Gain of EachInfluencing Factor In this study there were three factorsthat determined the purchase intentions of users A1 thestarting mode of the self-adaptive cruise A2 the position ofthe function key to adjust the distance between cars and A3the position of the function key of auto hold e proba-bilities of the factors and each condition are shown inTable 3

e information entropy values of the three factors wereas follows

E A1( 1113857 612

I(1 5) +612

I(4 2) 07842

E A2( 1113857 612

I(3 3) +612

I(2 4) 09592

E A3( 1113857 412

I(1 3) +412

I(1 3) +412

I(3 1) 08113

(8)

erefore the information gains of the factors were asfollows

Gain A1( 1113857 I(U) minus E A1( 1113857 01957

Gain A2( 1113857 I(U) minus E A2( 1113857 00207

Gain A3( 1113857 I(U) minus E A3( 1113857 01686

(9)

According to the information gains the starting mode ofthe self-adaptive cruise was the most important for the userrsquosperceived comfort and purchase intension while the posi-tion of the function key to adjust the distance between carswas not so significant By normalizing the information gainsof each influencing factor the results were taken as theweights of the factors which were

β1 Gain A1( 1113857

Gain A1( 1113857 + Gain A2( 1113857 + Gain A3( 1113857 05083

β2 Gain A2( 1113857

Gain A1( 1113857 + Gain A2( 1113857 + Gain A3( 1113857 00538

β3 Gain A3( 1113857

Gain A1( 1113857 + Gain A2( 1113857 + Gain A3( 1113857 04379

(10)

After taking these values into formula (1) the artificialfitness function could be constructed as follows

F 1

05083lowastQ1 + 00538lowastQ2 + 04379lowastQ3( 1113857 (11)

42 Results of GA

421 Characteristics of the Population e population sizeof this study was Pop 30 For the convenience of recordingthe unit of each dimension was decimeter (dm)e featuresof the 14 dimensions are described in Table 4

422 Solving Process of GA To match the userrsquos expecta-tion the seed of each initial population should be able to

Table 2 Classification of purchase intentions to interaction modes

e starting mode of self-adaptive cruise

e position of the function key to adjust thedistance between cars

e position of the function key ofauto-hold

Purchasedecision

Pulling the side stalk On the middle of the left spoke Under the EPB button 2Pressing the button On the middle of the left spoke Under the EPB button 2Pulling the side stalk On the bottom edge of the left spoke Under the EPB button 2Pressing the button On the bottom edge of the left spoke Under the EPB button 1Pulling the side stalk On the middle of the left spoke On the left of the EPB button 2Pressing the button On the middle of the left spoke On the left of the EPB button 1Pulling the side stalk On the bottom edge of the left spoke On the left of the EPB button 2Pressing the button On the bottom edge of the left spoke On the left of the EPB button 2Pulling the side stalk On the middle of the left spoke On the right of the EPB button 1Pressing the button On the middle of the left spoke On the right of the EPB button 1Pulling the side stalk On the bottom edge of the left spoke On the right of the EPB button 2Pressing the button On the bottom edge of the left spoke On the right of the EPB button 1

Table 3 e probabilities of the three factors and each condition

P (Vi) P (U1 | Vi) P (U2 | Vi)

A1Side stalk 612 16 56Button 612 46 26A2Middle 612 36 36Bottom edge 612 26 46A3Under the EPB button 412 14 34On the left of the EPB button 412 14 34On the right of the EPB button 412 34 14P (Vi) represents the probability of each condition of the influencingfactors P(U1 | Vi) and P(U2 | Vi) respectively indicate the conditionalprobabilities of the two categories

Complexity 7

cover part of the information of the PCD data And it shouldbe capable of satisfying the geometric description rules ofproduct evolution in terms of space and size [46] e initialpopulation of this paper came from repeated experiments inmany days and the perceptual experience and spatial at-tributes reflected by the data were stable e optimal so-lution of the model was obtained after 25 iterations Figure 3shows the evolutionary process of GA

In the figure the red line represents the best fitness valueof each generation And the blue dotted line describes themean fitness of each generation It could be seen from thefigure that the optimal value in the model increased with thenumber of iterations and converged into a straight line afterreaching the maximum value Until the optimal value of themodel remained unchanged the global optimal solutioncould be obtained

e genetic evolution stopped after 25 iterations It couldbe seen from the figure that although the deviation betweenthe optimal fitness values and the mean ones increased afterthe 9th iteration it continued to evolve which met theexpectation of evolution After the 21st iteration the devi-ation was close to 0

In the 25 generations of evolutionary iteration based onthe PCD data eight individuals with the highest fitness wereobtained which were the HMI dimension schemes that metthe usersrsquo expectation best in a certain generation of thepopulation It would be rational to make designs of HMIssuch as steering wheels and gear lever panels in the laterstage according to these fine values is could not onlyinherit the modelling genes of the HMIs but was alsocoupled to the user expectation to a certain extent e fi-nally generated dimensional data of the individuals with thehighest fitness are shown in Table 5

43 Significance of the Difference between the Original Di-mensions and the Evolved Ones To further validate whetherthe individuals with the highest fitness could bring about a

better take-over performance it was reasonable to collect theneeded time to complete driving assistant tasks under theHMI with the original sizes and the one with the evolvedsizes And the differences between the two sets of HMIdimensions could be analyzed

In the last of the 30 experiments we collected the time ofthe 13 subjects to complete the take-over tasks under theseven interaction modes in Table 1 Since the tasks had beenrepeated many times the subjects were skilled and the taskcompletion time would be relatively stable After making 3Dprinting models of the side stalk and the buttons accordingto the most suitable dimensions (Table 5) the drivingsimulator was reformeden the 13 subjects were invited torepeat the seven tasks with the new HMI and this set ofcompletion time was collected To judge whether the di-mensions could effectively improve the interaction perfor-mance and efficiency of driving information processingpaired-sample t-test was used on the two groups of data eresults are shown in Table 6

It could be seen from the results that except for A2 andC2 the completion time under the two sets of dimensions inthe process of the other five tasks all had a significant dif-ference (Figure 4)

Figure 4 shows that the completion time of each task wasshorter when the dimensions with high fitness were used onthe HMI is indicated that the driverrsquos information pro-cessing was faster which helped to improve the quality oftake-over and the ability to control the car As the driverrsquosattention to the HMI might compete with the execution ofdriving tasks for cognitive resources [47] and impair thedriving performance a shorter completion time of take-overtasks was conducive to the distribution of the driverrsquospsychological resources and more beneficial to trafficefficiency

It could be found by comparing the completion time ofeach task under the new HMI that (A) for starting the self-

Table 4 Features of the 14 dimensions

Tasks anddimensions

Mean(dm)

Stddeviation

Minimum(dm)

Maximum(dm)

A1 a11 19513 00789 18086 2091a12 06055 0063 05117 06977

A2 a21 08624 00242 08224 08979a22 00784 00477 00008 01487

B1 b11 13356 00633 12528 14475b12 00315 00169 00024 00588

B2 b21 13999 00526 13060 14952b22 0322 00139 03011 03497

C1 c11 29776 01254 28032 31940c12 3272 01414 30207 34928

C2 c21 27451 00305 27010 27966c22 32714 01304 30143 34819

C3 c31 32929 00596 32019 33963c32 3244 01523 30011 34932

2500735

0074

00745

0075

00755

0076

00765

0077

00775

0078

Fittn

ess

5 10 15 200Generation

Mean fitnessBest fitness

Figure 3 Solving process of GA

8 Complexity

adaptive cruise the completion time of operating with a sidestalk was shorter (B) for the task of adjusting the distancebetween cars the completion time would be shorter whenoperating with a button located in the middle of the steeringwheel spoke (C) for the task of auto hold a starting buttonon the left side of the EPB button could bring about a shortercompletion time ese results provided a reference for thefuture automobile interior design

5 Discussion

Driving assistant system (DAS) plays an important role inthe context of human-computer codriving As reportedDAS can help the driver to complete the intense drivingtasks effectively and abate the pressure level of the driver[48] erefore it is meaningful to make a study for thecontroller dimensions in the DAS e study collectedcomfortable dimensions perceived by the drivers in a relaxedstate e aim was to detect a set of reasonable dimensions

which could make drivers at a low level of stress In ad-dition it is valuable for drivers to adapt to the different roadenvironment and realize sustainable traffic to optimize thesizes of interactive components on the in-vehicle HMI emain goals of in-vehicle technologies and cooperativeservices were to reduce traffic congestion and improve roadsafety [4]

e study attempted to analyze the optimal HMI di-mensions suitable for take-over tasks e involved HMIcomponents such as the steering wheel self-adaptive cruiselever distance adjustment button and auto hold buttonwere all direct design elements for users to see and touchBecause the design and performance of a product wereimportant factors affecting the decision-making of con-sumers [49] we calculated the information entropy andinformation gains of the three kinds of take-over tasks forclassifying usersrsquo purchase intentions e informationtransmitted by a product was multivariant complex andfuzzy so that there would be dissipation in the process oftransmission erefore the image evaluation process ofproducts could be regarded as a chaotic state e feedbackinformation could be calculated by information entropy tocarry out a comprehensive evaluation e smaller the en-tropy of an image the clearer it was From the perspective ofchaos degree the 2nd factor in Table 3 (the position of thefunction key to adjust the distance between cars) broughtmore uncertainty to the purchase decision(E(A2) 09592) which indicated that little informationwas provided by the 2nd factor for judging the usersrsquo de-cisions On the contrary the uncertainty produced by the 1stfactor (the starting mode of the self-adaptive cruise) was thesmallest (E(A1) 07842) is illustrated that using theside stalk or a button to complete the switch of the controlright had a great impact on the satisfaction and purchasedecision of users

After normalizing the information entropy of threekinds of take-over tasks their weight values were calculated

Table 5 e finally generated dimensional data of the individuals with the highest fitness

(A) e starting mode of self-adaptive cruise

(B) e position of thefunction key to adjust thedistance between cars

(C) e position of the function key of auto-hold Fitness value

a11 a12 a21 a22 b11 b12 b21 b22 c11 c12 c21 c22 c31 c32202 062 089 009 136 004 148 030 284 302 271 338 321 313 008

Table 6 e paired sample t-test results of task completion time under the two HMIs with different dimensions

Interaction modes of take-over tasks

e task completiontime under the original

dimensions (s)

e task completiontime under the evolved

dimensions (s) t-value Sig

Mean Std dev Mean Std devA1lowast 0715 0185 0537 0167 2553 0025A2 0657 0181 0637 0186 1151 0272B1lowast 0648 013 0462 0206 2797 0016B2lowastlowast 0736 0106 0571 0124 3116 0009C1lowast 0753 0295 0562 0222 2515 0027C2 0667 0226 0522 0176 1999 0069C3lowast 069 0208 0537 0225 2355 0036

A1lowast B1lowast B2lowastlowast C1lowast C3lowast

0

02

04

06

08

1

12

Time under the original dimensionsTime under the evolved dimensions

Figure 4 e completion time of the tasks with significant dif-ferences under the two HMIs with different dimensions

Complexity 9

and the fitness function was constructed In this way thedimensions with the highest fitness could be found out bymeans of GA

e background and driving experience of the subjects weredifferent and their familiarity with the take-over tasks and theHMI was variant Besides the noise of usersrsquo subjective per-ception data was always high Taking the above factors intoconsideration we collected PCD data repeatedly from the same13 subjects to form the initial population From the evolutionarytrajectory it could be found that in the 8th 10th 11th 14th 17thand 21st iteration there appeared an optimal solution

GA was always used for parameter optimization bysearching the optimal solutions while seldom used in userresearch and environment perception analysis But in studiesof user-centred design it could search the set of satisfactorysolutions driven by user expectations [46] Many pieces ofresearch had utilized GA to improve the ergonomic effi-ciency or optimize interactive component sizes in an op-erating environment [34ndash38 50 51] In this study wereported a driver perception research which investigated themost comfortable HMI sizes under different interactionmodes of take-over tasks By means of normalized geometricselection arithmetic uniform crossover and nonuniformmutation GA was applied in the research of perceivedcomfortable dimensions and a set of reasonable dimensionswas obtained e dimensional evolutionary method dis-cussed in this study embodied the concept of dynamic in-teraction design which took the mutual influence andrestrictions among different dimensions into considerationCompared with traditional methods of measuring manytimes and taking the average this method was more ad-vanced And the study enriched theoretical researches in thefield of product design and collaborative optimization

Perceived comfort was a complex problem for its dif-ficulty to be quantified objectively But it was useful in er-gonomic explorations Researchers had found that theperceptual messages from product property and usage re-quirements were competent to the analysis of user experi-ence [52] erefore it was reasonable to probe suitabledimensions by perceived comfort For the study of HMIdesign effect in a specific environment these data couldproduce important and dependable results [53] Rationaldimensions of the automobile interior HMI played an im-portant role in the driverrsquos operation quality Caruso [54]studied driversrsquo experience on the seat by an experiment andfound that interior space influenced the driversrsquo ergonomicperformance [54] For the take-over tasks in the situation ofhuman-computer codriving the dimensions also affectedwhether the driver could safely achieve the control rightswitch in the shortest possible time

In this paper paired-sample t-test was used to testwhether the sizes with the highest fitness could bring about asignificant reduction in task completion time In the resultswe found that when the interface adopted the finally evolvedsizes the completion time of each task was shorter As thecompletion time of in-vehicle interactive tasks was related tothe vehicle speed and driver errors under the context ofhuman-computer codriving [21 55] a shorter completiontime would help the driver to control the speed better

6 Limitation

e study explored the reasonable dimensions of the hu-man-machine interface in intelligent cars Neverthelessthere were still some deficiencies in the equipment used inthis study e driving simulator was easier to control andreform and helped to reduce the subjectsrsquo pressure How-ever the immersion visual interactivity and fidelity degreewere not high enough After obtaining reliable conclusionsin this study outdoor experiment with a real vehicle could beconducted in the next stage

Besides the sample size of this study was limited Whenthe sample size was larger in future studies the represen-tativeness of the data would be better and the results of GAcould be further improved

Finally usersrsquo perception of the HMI dimensions de-viated from each other In the next phase of the studyelectroencephalography (EEG) signal in the process ofperforming the driving tasks could be used By this meansthe subjectsrsquo attention and intention when interacting withthe HMI could be reflected in an objective way [56 57] as asupplement to their subjective report

7 Conclusion

For HMI design of industrial products physical concretemodelling parameters such as the product shape size andlayout were the objective properties that users could directlyperceive and the basis for the product to be expressed Basedon the dynamic measurement principle this paper proposedan optimized design method for ergonomic dimensions byGA e results indicated that the method was effective andprovided a new way for product shape innovation andhuman-machine dimensional automatic design e mainconclusions of this study were as follows

(1) Based on the information entropy theory the weightvalues of the three driving assistance functions werecalculated e results showed that the informationgain of the starting mode of the self-adaptive cruisewas highest while the information gain of the po-sition of the function key to adjust the distancebetween cars was lowest which provided a theo-retical basis for the dimensional fitness evaluationmechanism of the HMI

(2) e evolutionary process of HMI dimensionsdriven by user expectation was studied and theevolution conditions were analyzed e studyconfirmed the feasibility of normalized geometricselection arithmetic uniform crossover and non-uniform mutation in iterative researches for hu-man-machine dimensions rough parameterdebugging this evolutionary method was proved tobe effective in obtaining appropriate sizes of HMIcomponents e model could make full use of theadvantages of GA to improve the user researchperformance such as few restrictions on the opti-mization high robustness and an easily achievedglobal search

10 Complexity

(3) After finding the dimensions with the highest fitnessand applying them to the design of the HMI within-subjects experiments showed that the completiontime of five take-over tasks was significantly shorterexcept for that of Task A2 and C2is indicated thatthe information processing of drivers under the HMIwith evolved dimensions was faster which was morebeneficial to the vehicle control and take-overquality Intelligent vehicles achieved intelligent in-formation exchange with people vehicles and roadsthrough in-vehicle sensor system and informationterminals erefore the evolved dimensions couldmake the vehicle apt to respond to the environmentand have more time to fully deal with the complextraffic environment and would help to improve thetraffic efficiency therefrom

Data Availability

e data used to support the findings of this study areavailable from the corresponding author upon request

Consent

e study respected and guaranteed the subjectsrsquo right toparticipate in the study and allowed them to withdraw from thestudy unconditionally at any stage e study strictly followedthe informed consent procedure e subjects were asked tosign a consent form to inform them of the purpose processduration of the experiment matters needing attention and themethods and purpose of collecting the data so as to reducetheir anxiety and confusion In the experiment the personalsafety and health of the subjects took priority According to theusage guidelines of the driving simulator we tried to avoid anypossible injury in driving tasks e privacy of the subjects wasstrictly protected e subjects were informed of the storageuse and confidentiality measures of their personal informationand image data in advance truthfully And we would notdisclose the information to a third party or spread them on theInternet without authorization

Conflicts of Interest

e authors declare that they have no conflicts of interest

Acknowledgments

is researchwas funded by BeijingUrbanGovernance ResearchProject (grant no 20XN244) Scientific Research Program ofBeijing Education Commission (grant no KM202010009003)Yuyou Talent Support Program of North China University ofTechnology (grant no 107051360018XN012018) Tianjin YouthTalents Reserve Support Program and Chinese ErgonomicsSociety amp Kingfar Joint Research Fund for Outstanding YoungScholars (grant number CES-Kingfar-2019-001)

References

[1] S Jin J Yang E Wang and J Liu ldquoe influence of high-speed rail on ice-snow tourism in north eastern ChinardquoTourism Management vol 78 Article ID 104070 2020

[2] J Yang R Yang J Sun T Huang and Q Ge ldquoe spatialdifferentiation of the suitability of ice-snow tourist destina-tions based on a comprehensive evaluation model in ChinardquoSustainability vol 9 no 5 Article ID 774 2017

[3] W-Y Chung T-W Chong and B-G Lee ldquoMethods to detectand reduce driver stress a reviewrdquo International Journal ofAutomotive Technology vol 20 no 5 pp 1051ndash1063 2019

[4] H Farah and H N Koutsopoulos ldquoDo cooperative systemsmake driversrsquo car-following behavior saferrdquo TransportationResearch Part C Emerging Technologies vol 41 pp 61ndash722014

[5] SAE Taxonomy and Definitions for Terms Related to DrivingAutomation Systems for On-Road Motor Vehicles SAEJ3016TM Warrendale PA USA 2018

[6] J Hoegg J W Alba and D W Dahl ldquoe good the bad andthe ugly influence of aesthetics on product feature judg-mentsrdquo Journal of Consumer Psychology vol 20 no 4pp 419ndash430 2010

[7] M Kyriakidis R Happee and J C F De Winter ldquoPublicopinion on automated driving results of an internationalquestionnaire among 5000 respondentsrdquo TransportationResearch Part F Traffic Psychology and Behaviour vol 32pp 127ndash140 2015

[8] F Meng and C Spence ldquoTactile warning signals for in-vehiclesystemsrdquo Accident Analysis amp Prevention vol 75 pp 333ndash346 2015

[9] J Wan and C Wu ldquoe effects of lead time of take-overrequest and nondriving tasks on taking-over control of au-tomated vehiclesrdquo IEEE Transactions on Human MachineSystems vol 48 no 6 pp 582ndash591 2018

[10] S Petermeijer P Hornberger I Ganotis J D Winter andK Bengler ldquoe design of a vibrotactile seat for conveyingtake-over requests in automated drivingrdquo in Proceedings of the8th International Conference on Applied Human Factors andErgonomics Los Angeles CA USA July 2017

[11] T Ito A Takata and K Oosawa ldquoTime required for take-overfrom automated to manual drivingrdquo in Proceedings of the SAE2016 World Congress and Exhibition Detroit MI USA April2016

[12] K Zeeb A Buchner and M Schrauf ldquoWhat determines thetake-over time An integrated model approach of driver take-over after automated drivingrdquo Accident Analysis amp Preven-tion vol 78 pp 212ndash221 2015

[13] B Mok M Johns K J Lee et al ldquoEmergency automation offunstructured transition timing for distracted drivers of au-tomated vehiclesrdquo in Proceedings of the IEEE InternationalConference on Intelligent Transportation Systems Las PalmasSpain September 2015

[14] S Samuel A Borowsky S Zilberstein and D L FisherldquoMinimum time to situation awareness in scenarios involvingtransfer of control from an automated driving suiterdquoTransportation Research Record Journal of the TransportationResearch Board vol 2602 no 2602 pp 115ndash120 2016

[15] W Vlakveld N van Nes J de Bruin L Vissers andM van der Kroft ldquoSituation awareness increases when drivershave more time to take over the wheel in a level 3 automated

Complexity 11

car a simulator studyrdquo Transportation Research Part F TrafficPsychology and Behaviour vol 58 pp 917ndash929 2018

[16] N Merat A H Jamson F C H Lai M Daly andO M J Carsten ldquoTransition to manual driver behaviourwhen resuming control from a highly automated vehiclerdquoTransportation Research Part F Traffic Psychology and Be-haviour vol 27 no 26 pp 274ndash282 2014

[17] C Z Wu H R Wu and N C Lyu ldquoReview of control switchand safety of human-computer driving intelligent vehiclerdquoJournal of Traffic and Transportation Engineering vol 18no 6 pp 131ndash141 2018

[18] C Gold D Dambock L Lorenz and K Bengler ldquoldquoTakeoverrdquo how long does it take to get the driver back into thelooprdquo Proceedings of the Human Factors and ErgonomicsSociety Annual Meeting vol 57 no 1 pp 1938ndash1942 2013

[19] H Yang Y Zhao and Y Wang ldquoIdentifying modeling formsof instrument panel system in intelligent shared cars a studyfor perceptual preference and in-vehicle behaviorsrdquo Envi-ronmental Science and Pollution Research vol 27 no 1pp 1009ndash1023 2020

[20] A Berkovich A Lu B Levine and A V Reddy ldquoObservedcustomer seating and standing behavior and seat preferenceson board subway cars in New York cityrdquo TransportationResearch Record Journal of the Transportation ResearchBoard vol 2353 no 1 pp 33ndash46 2013

[21] R Li Y V Chen C Sha and Z Lu ldquoEffects of interface layouton the usability of in-vehicle information systems and drivingsafetyrdquo Displays vol 49 pp 124ndash132 2017

[22] H Kim S Kwon J Heo H Lee and M K Chung ldquoe effectof touch-key size on the usability of in-vehicle informationsystems and driving safety during simulated drivingrdquo AppliedErgonomics vol 45 no 3 pp 379ndash388 2014

[23] H Kim and H Song ldquoEvaluation of the safety and usability oftouch gestures in operating in-vehicle information systemswith visual occlusionrdquo Applied Ergonomics vol 45 no 3pp 789ndash798 2014

[24] M Koerber C Gold D Lechner and K Bengler ldquoe in-fluence of age on the take-over of vehicle control in highlyautomated drivingrdquo Transportation Research Part F TrafficPsychology amp Behaviour vol 39 pp 19ndash32 2016

[25] H Clark A C Mclaughlin B Williams and J Feng ldquoPer-formance in takeover and characteristics of non-driving re-lated tasks during highly automated driving in younger andolder driversrdquo Proceedings of the Human Factors amp Ergo-nomics Society Annual Meeting vol 61 no 1 pp 37ndash41 2017

[26] I Lijarcio S A Useche J Llamazares and L MontoroldquoPerceived benefits and constraints in vehicle automationdata to assess the relationship between driverrsquos features andtheir attitudes towards autonomous vehiclesrdquo Data in Briefvol 27 Article ID 104662 2019

[27] K-Z Chen and X-A Feng ldquoVirtual genes of manufacturingproducts and their reforms for product innovative designrdquoProceedings of the Institution of Mechanical Engineers Part CJournal of Mechanical Engineering Science vol 218 no 5pp 557ndash574 2004

[28] K-Z Chen X-A Feng and X-C Chen ldquoReverse deductionof virtual chromosomes of manufactured products for theirgene-engineering-based innovative designrdquo Computer-AidedDesign vol 37 no 11 pp 1191ndash1203 2005

[29] H-C Tsai and J-R Chou ldquoAutomatic design support andimage evaluation of two-coloured products using colour as-sociation and colour harmony scales and genetic algorithmrdquoComputer-Aided Design vol 39 no 9 pp 818ndash828 2007

[30] S-W Hsiao C-F Hsu and K-W Tang ldquoA consultation andsimulation system for product color planning based on in-teractive genetic algorithmsrdquo Color Research amp Applicationvol 38 no 5 pp 375ndash390 2013

[31] Y Sun W-l Wang Y-w Zhao and X-j Liu ldquoUser imageoriented interactive genetic algorithm evaluation mode inproduct developmentrdquo Computer Integrated ManufacturingSystems vol 18 no 2 pp 276ndash281 2012

[32] R Kamalian E Yeh Z Ying A M Agogino and H TakagildquoReducing human fatigue in interactive evolutionary com-putation through fuzzy systems and machine learning sys-temsrdquo in Proceedings of the 2006 IEEE InternationalConference on Fuzzy Systems Vancouver Canada July 2006

[33] Z Gu M Xi Tang and J H Frazer ldquoCapturing aestheticintention during interactive evolutionrdquo Computer-AidedDesign vol 38 no 3 pp 224ndash237 2006

[34] G Harih M Borovinsek and Z Ren Optimisation ofProductrsquos Hand-Handle Interface Material Parameters forImproved Ergonomics Springer International PublishingCham Switzerland 2015

[35] P Cheng D Chen and J Wang ldquoClustering of the bodyshape of the adult male by using principal component analysisand genetic algorithm-BP neural networkrdquo Soft Computingvol 24 no 17 Article ID 13219 2020

[36] N M B Serrano P J G Nieto A S Sanchez F S Lasherasand P R Fernandez AHybrid Algorithm for the Assessment ofthe Influence of Risk Factors in the Development of Upper LimbMusculoskeletal Disorders Springer Cham Switzerland 2018

[37] S S Sankar O-M Holman A F Gazabon and C J AcevedoldquoApplication of genetic algorithm to job scheduling underergonomic constraints in manufacturing industryrdquo Journal ofAmbient Intelligence and Humanized Computing vol 10no 5 pp 2063ndash2090 2019

[38] A-Z Atiya L Lee and K Xing ldquoDeveloping a multi-ob-jective genetic optimisation approach for an operationaldesign of a manual mixed-model assembly line with walkingworkersrdquo Journal of Intelligent Manufacturing vol 27 no 5pp 1ndash17 2014

[39] P Jedlicka and T Ryba ldquoGenetic algorithm application inimage segmentationrdquo Pattern Recognition and Image Anal-ysis vol 26 no 3 pp 497ndash501 2016

[40] L V Campenhout J Frens K Overbeeke A Standaert andH Peremans ldquoPhysical interaction in a dematerializedworldrdquo International Journal of Design vol 7 no 1 pp 1ndash182013

[41] K Nandhini and S R Balasundaram ldquoImproving readabilitythrough individualized summary extraction using interactivegenetic algorithmrdquo Applied Artificial Intelligence vol 30no 7 pp 635ndash661 2016

[42] R M Gray Entropy and Information 6eory Springer NewYork NY USA 2nd edition 2011

[43] B Anderhofstadt and S Spinler ldquoFactors affecting the pur-chasing decision and operation of alternative fuel-poweredheavy-duty trucks in Germanymdasha Delphi studyrdquo Trans-portation Research Part D Transport and Environmentvol 73 pp 87ndash107 2019

[44] X-m Wu and Y Feng ldquoReactive power optimization ofdistribution network based on improved genetic algorithmrdquoJournal of Xirsquoan Shiyou University (Natural Science Edition)vol 30 no 3 pp 95ndash99 2015

[45] Z Ling andM-j Li ldquoState space evolutionary algorithm basedon non-uniform mutation operatorrdquo Computer Technologyand Development vol 28 no 9 pp 68ndash71 2018

12 Complexity

[46] W Hu and J Zhao ldquoAutomobile styling gene evolutiondriven by usersrsquo expectation imagerdquo Journal of MechanicalEngineering vol 47 no 16 pp 176ndash181 2011

[47] J J Scott and R Gray ldquoA comparison of tactile visual andauditory warnings for rear-end collision prevention in sim-ulated drivingrdquo Human Factors 6e Journal of the HumanFactors and Ergonomics Society vol 50 no 2 pp 264ndash2752008

[48] B Reimer B Mehler and J F Coughlin ldquoReductions in self-reported stress and anticipatory heart rate with the use of asemi-automated parallel parking systemrdquo Applied Ergonom-ics vol 52 pp 120ndash127 2016

[49] B DrsquoIppolito ldquoe importance of design for firms com-petitiveness a review of the literaturerdquo Technovation vol 34no 11 pp 716ndash730 2014

[50] S Asensio-Cuesta J A Diego-Mas L Canos-Daros andC Andres-Romano ldquoA genetic algorithm for the design of jobrotation schedules considering ergonomic and competencecriteriardquo International Journal of Advanced ManufacturingTechnology vol 60 no 9ndash12 pp 1161ndash1174 2012

[51] S-C Lee H-E Tseng C-C Chang and Y-M HuangldquoApplying interactive genetic algorithms to disassembly se-quence planningrdquo International Journal of Precision Engi-neering and Manufacturing vol 21 no 4 pp 663ndash679 2020

[52] M G Shishaev V V Dikovitsky and L V Lapochkina ldquoeexperience of building cognitive user interfaces of multido-main information systems based on the mental model ofusersrdquo in Proceedings of the 6th Computer Science On-LineConference (CSOC 2017) Prague Czech Republic April 2017

[53] S Lee H Alzoubi and S Kim ldquoe effect of interior designelements and lighting layouts on prospective occupantsrsquoperceptions of amenity and efficiency in living roomsrdquo Sus-tainability vol 9 no 7 Article ID 1119 2017

[54] G Caruso ldquoMixed reality system for ergonomic assessment ofdriverrsquos seatrdquo International Journal of Virtual Reality vol 10no 2 pp 69ndash79 2011

[55] M L Reyes and J D Lee ldquoEffects of cognitive load presenceand duration on driver eye movements and event detectionperformancerdquo Transportation Research Part F Traffic Psy-chology and Behaviour vol 11 no 6 pp 391ndash402 2008

[56] D Cisler P M Greenwood D M Roberts R McKendrickand C L Baldwin ldquoComparing the relative strengths of EEGand low-cost physiological devices in modeling attentionallocation in semiautonomous vehiclesrdquo Frontiers in HumanNeuroscience vol 13 Article ID 109 2019

[57] I-H Kim J-W Kim S Haufe and S-W Lee ldquoDetection ofbraking intention in diverse situations during simulateddriving based on EEG feature combinationrdquo Journal of NeuralEngineering vol 12 no 1 Article ID 016001 2014

Complexity 13

cover part of the information of the PCD data And it shouldbe capable of satisfying the geometric description rules ofproduct evolution in terms of space and size [46] e initialpopulation of this paper came from repeated experiments inmany days and the perceptual experience and spatial at-tributes reflected by the data were stable e optimal so-lution of the model was obtained after 25 iterations Figure 3shows the evolutionary process of GA

In the figure the red line represents the best fitness valueof each generation And the blue dotted line describes themean fitness of each generation It could be seen from thefigure that the optimal value in the model increased with thenumber of iterations and converged into a straight line afterreaching the maximum value Until the optimal value of themodel remained unchanged the global optimal solutioncould be obtained

e genetic evolution stopped after 25 iterations It couldbe seen from the figure that although the deviation betweenthe optimal fitness values and the mean ones increased afterthe 9th iteration it continued to evolve which met theexpectation of evolution After the 21st iteration the devi-ation was close to 0

In the 25 generations of evolutionary iteration based onthe PCD data eight individuals with the highest fitness wereobtained which were the HMI dimension schemes that metthe usersrsquo expectation best in a certain generation of thepopulation It would be rational to make designs of HMIssuch as steering wheels and gear lever panels in the laterstage according to these fine values is could not onlyinherit the modelling genes of the HMIs but was alsocoupled to the user expectation to a certain extent e fi-nally generated dimensional data of the individuals with thehighest fitness are shown in Table 5

43 Significance of the Difference between the Original Di-mensions and the Evolved Ones To further validate whetherthe individuals with the highest fitness could bring about a

better take-over performance it was reasonable to collect theneeded time to complete driving assistant tasks under theHMI with the original sizes and the one with the evolvedsizes And the differences between the two sets of HMIdimensions could be analyzed

In the last of the 30 experiments we collected the time ofthe 13 subjects to complete the take-over tasks under theseven interaction modes in Table 1 Since the tasks had beenrepeated many times the subjects were skilled and the taskcompletion time would be relatively stable After making 3Dprinting models of the side stalk and the buttons accordingto the most suitable dimensions (Table 5) the drivingsimulator was reformeden the 13 subjects were invited torepeat the seven tasks with the new HMI and this set ofcompletion time was collected To judge whether the di-mensions could effectively improve the interaction perfor-mance and efficiency of driving information processingpaired-sample t-test was used on the two groups of data eresults are shown in Table 6

It could be seen from the results that except for A2 andC2 the completion time under the two sets of dimensions inthe process of the other five tasks all had a significant dif-ference (Figure 4)

Figure 4 shows that the completion time of each task wasshorter when the dimensions with high fitness were used onthe HMI is indicated that the driverrsquos information pro-cessing was faster which helped to improve the quality oftake-over and the ability to control the car As the driverrsquosattention to the HMI might compete with the execution ofdriving tasks for cognitive resources [47] and impair thedriving performance a shorter completion time of take-overtasks was conducive to the distribution of the driverrsquospsychological resources and more beneficial to trafficefficiency

It could be found by comparing the completion time ofeach task under the new HMI that (A) for starting the self-

Table 4 Features of the 14 dimensions

Tasks anddimensions

Mean(dm)

Stddeviation

Minimum(dm)

Maximum(dm)

A1 a11 19513 00789 18086 2091a12 06055 0063 05117 06977

A2 a21 08624 00242 08224 08979a22 00784 00477 00008 01487

B1 b11 13356 00633 12528 14475b12 00315 00169 00024 00588

B2 b21 13999 00526 13060 14952b22 0322 00139 03011 03497

C1 c11 29776 01254 28032 31940c12 3272 01414 30207 34928

C2 c21 27451 00305 27010 27966c22 32714 01304 30143 34819

C3 c31 32929 00596 32019 33963c32 3244 01523 30011 34932

2500735

0074

00745

0075

00755

0076

00765

0077

00775

0078

Fittn

ess

5 10 15 200Generation

Mean fitnessBest fitness

Figure 3 Solving process of GA

8 Complexity

adaptive cruise the completion time of operating with a sidestalk was shorter (B) for the task of adjusting the distancebetween cars the completion time would be shorter whenoperating with a button located in the middle of the steeringwheel spoke (C) for the task of auto hold a starting buttonon the left side of the EPB button could bring about a shortercompletion time ese results provided a reference for thefuture automobile interior design

5 Discussion

Driving assistant system (DAS) plays an important role inthe context of human-computer codriving As reportedDAS can help the driver to complete the intense drivingtasks effectively and abate the pressure level of the driver[48] erefore it is meaningful to make a study for thecontroller dimensions in the DAS e study collectedcomfortable dimensions perceived by the drivers in a relaxedstate e aim was to detect a set of reasonable dimensions

which could make drivers at a low level of stress In ad-dition it is valuable for drivers to adapt to the different roadenvironment and realize sustainable traffic to optimize thesizes of interactive components on the in-vehicle HMI emain goals of in-vehicle technologies and cooperativeservices were to reduce traffic congestion and improve roadsafety [4]

e study attempted to analyze the optimal HMI di-mensions suitable for take-over tasks e involved HMIcomponents such as the steering wheel self-adaptive cruiselever distance adjustment button and auto hold buttonwere all direct design elements for users to see and touchBecause the design and performance of a product wereimportant factors affecting the decision-making of con-sumers [49] we calculated the information entropy andinformation gains of the three kinds of take-over tasks forclassifying usersrsquo purchase intentions e informationtransmitted by a product was multivariant complex andfuzzy so that there would be dissipation in the process oftransmission erefore the image evaluation process ofproducts could be regarded as a chaotic state e feedbackinformation could be calculated by information entropy tocarry out a comprehensive evaluation e smaller the en-tropy of an image the clearer it was From the perspective ofchaos degree the 2nd factor in Table 3 (the position of thefunction key to adjust the distance between cars) broughtmore uncertainty to the purchase decision(E(A2) 09592) which indicated that little informationwas provided by the 2nd factor for judging the usersrsquo de-cisions On the contrary the uncertainty produced by the 1stfactor (the starting mode of the self-adaptive cruise) was thesmallest (E(A1) 07842) is illustrated that using theside stalk or a button to complete the switch of the controlright had a great impact on the satisfaction and purchasedecision of users

After normalizing the information entropy of threekinds of take-over tasks their weight values were calculated

Table 5 e finally generated dimensional data of the individuals with the highest fitness

(A) e starting mode of self-adaptive cruise

(B) e position of thefunction key to adjust thedistance between cars

(C) e position of the function key of auto-hold Fitness value

a11 a12 a21 a22 b11 b12 b21 b22 c11 c12 c21 c22 c31 c32202 062 089 009 136 004 148 030 284 302 271 338 321 313 008

Table 6 e paired sample t-test results of task completion time under the two HMIs with different dimensions

Interaction modes of take-over tasks

e task completiontime under the original

dimensions (s)

e task completiontime under the evolved

dimensions (s) t-value Sig

Mean Std dev Mean Std devA1lowast 0715 0185 0537 0167 2553 0025A2 0657 0181 0637 0186 1151 0272B1lowast 0648 013 0462 0206 2797 0016B2lowastlowast 0736 0106 0571 0124 3116 0009C1lowast 0753 0295 0562 0222 2515 0027C2 0667 0226 0522 0176 1999 0069C3lowast 069 0208 0537 0225 2355 0036

A1lowast B1lowast B2lowastlowast C1lowast C3lowast

0

02

04

06

08

1

12

Time under the original dimensionsTime under the evolved dimensions

Figure 4 e completion time of the tasks with significant dif-ferences under the two HMIs with different dimensions

Complexity 9

and the fitness function was constructed In this way thedimensions with the highest fitness could be found out bymeans of GA

e background and driving experience of the subjects weredifferent and their familiarity with the take-over tasks and theHMI was variant Besides the noise of usersrsquo subjective per-ception data was always high Taking the above factors intoconsideration we collected PCD data repeatedly from the same13 subjects to form the initial population From the evolutionarytrajectory it could be found that in the 8th 10th 11th 14th 17thand 21st iteration there appeared an optimal solution

GA was always used for parameter optimization bysearching the optimal solutions while seldom used in userresearch and environment perception analysis But in studiesof user-centred design it could search the set of satisfactorysolutions driven by user expectations [46] Many pieces ofresearch had utilized GA to improve the ergonomic effi-ciency or optimize interactive component sizes in an op-erating environment [34ndash38 50 51] In this study wereported a driver perception research which investigated themost comfortable HMI sizes under different interactionmodes of take-over tasks By means of normalized geometricselection arithmetic uniform crossover and nonuniformmutation GA was applied in the research of perceivedcomfortable dimensions and a set of reasonable dimensionswas obtained e dimensional evolutionary method dis-cussed in this study embodied the concept of dynamic in-teraction design which took the mutual influence andrestrictions among different dimensions into considerationCompared with traditional methods of measuring manytimes and taking the average this method was more ad-vanced And the study enriched theoretical researches in thefield of product design and collaborative optimization

Perceived comfort was a complex problem for its dif-ficulty to be quantified objectively But it was useful in er-gonomic explorations Researchers had found that theperceptual messages from product property and usage re-quirements were competent to the analysis of user experi-ence [52] erefore it was reasonable to probe suitabledimensions by perceived comfort For the study of HMIdesign effect in a specific environment these data couldproduce important and dependable results [53] Rationaldimensions of the automobile interior HMI played an im-portant role in the driverrsquos operation quality Caruso [54]studied driversrsquo experience on the seat by an experiment andfound that interior space influenced the driversrsquo ergonomicperformance [54] For the take-over tasks in the situation ofhuman-computer codriving the dimensions also affectedwhether the driver could safely achieve the control rightswitch in the shortest possible time

In this paper paired-sample t-test was used to testwhether the sizes with the highest fitness could bring about asignificant reduction in task completion time In the resultswe found that when the interface adopted the finally evolvedsizes the completion time of each task was shorter As thecompletion time of in-vehicle interactive tasks was related tothe vehicle speed and driver errors under the context ofhuman-computer codriving [21 55] a shorter completiontime would help the driver to control the speed better

6 Limitation

e study explored the reasonable dimensions of the hu-man-machine interface in intelligent cars Neverthelessthere were still some deficiencies in the equipment used inthis study e driving simulator was easier to control andreform and helped to reduce the subjectsrsquo pressure How-ever the immersion visual interactivity and fidelity degreewere not high enough After obtaining reliable conclusionsin this study outdoor experiment with a real vehicle could beconducted in the next stage

Besides the sample size of this study was limited Whenthe sample size was larger in future studies the represen-tativeness of the data would be better and the results of GAcould be further improved

Finally usersrsquo perception of the HMI dimensions de-viated from each other In the next phase of the studyelectroencephalography (EEG) signal in the process ofperforming the driving tasks could be used By this meansthe subjectsrsquo attention and intention when interacting withthe HMI could be reflected in an objective way [56 57] as asupplement to their subjective report

7 Conclusion

For HMI design of industrial products physical concretemodelling parameters such as the product shape size andlayout were the objective properties that users could directlyperceive and the basis for the product to be expressed Basedon the dynamic measurement principle this paper proposedan optimized design method for ergonomic dimensions byGA e results indicated that the method was effective andprovided a new way for product shape innovation andhuman-machine dimensional automatic design e mainconclusions of this study were as follows

(1) Based on the information entropy theory the weightvalues of the three driving assistance functions werecalculated e results showed that the informationgain of the starting mode of the self-adaptive cruisewas highest while the information gain of the po-sition of the function key to adjust the distancebetween cars was lowest which provided a theo-retical basis for the dimensional fitness evaluationmechanism of the HMI

(2) e evolutionary process of HMI dimensionsdriven by user expectation was studied and theevolution conditions were analyzed e studyconfirmed the feasibility of normalized geometricselection arithmetic uniform crossover and non-uniform mutation in iterative researches for hu-man-machine dimensions rough parameterdebugging this evolutionary method was proved tobe effective in obtaining appropriate sizes of HMIcomponents e model could make full use of theadvantages of GA to improve the user researchperformance such as few restrictions on the opti-mization high robustness and an easily achievedglobal search

10 Complexity

(3) After finding the dimensions with the highest fitnessand applying them to the design of the HMI within-subjects experiments showed that the completiontime of five take-over tasks was significantly shorterexcept for that of Task A2 and C2is indicated thatthe information processing of drivers under the HMIwith evolved dimensions was faster which was morebeneficial to the vehicle control and take-overquality Intelligent vehicles achieved intelligent in-formation exchange with people vehicles and roadsthrough in-vehicle sensor system and informationterminals erefore the evolved dimensions couldmake the vehicle apt to respond to the environmentand have more time to fully deal with the complextraffic environment and would help to improve thetraffic efficiency therefrom

Data Availability

e data used to support the findings of this study areavailable from the corresponding author upon request

Consent

e study respected and guaranteed the subjectsrsquo right toparticipate in the study and allowed them to withdraw from thestudy unconditionally at any stage e study strictly followedthe informed consent procedure e subjects were asked tosign a consent form to inform them of the purpose processduration of the experiment matters needing attention and themethods and purpose of collecting the data so as to reducetheir anxiety and confusion In the experiment the personalsafety and health of the subjects took priority According to theusage guidelines of the driving simulator we tried to avoid anypossible injury in driving tasks e privacy of the subjects wasstrictly protected e subjects were informed of the storageuse and confidentiality measures of their personal informationand image data in advance truthfully And we would notdisclose the information to a third party or spread them on theInternet without authorization

Conflicts of Interest

e authors declare that they have no conflicts of interest

Acknowledgments

is researchwas funded by BeijingUrbanGovernance ResearchProject (grant no 20XN244) Scientific Research Program ofBeijing Education Commission (grant no KM202010009003)Yuyou Talent Support Program of North China University ofTechnology (grant no 107051360018XN012018) Tianjin YouthTalents Reserve Support Program and Chinese ErgonomicsSociety amp Kingfar Joint Research Fund for Outstanding YoungScholars (grant number CES-Kingfar-2019-001)

References

[1] S Jin J Yang E Wang and J Liu ldquoe influence of high-speed rail on ice-snow tourism in north eastern ChinardquoTourism Management vol 78 Article ID 104070 2020

[2] J Yang R Yang J Sun T Huang and Q Ge ldquoe spatialdifferentiation of the suitability of ice-snow tourist destina-tions based on a comprehensive evaluation model in ChinardquoSustainability vol 9 no 5 Article ID 774 2017

[3] W-Y Chung T-W Chong and B-G Lee ldquoMethods to detectand reduce driver stress a reviewrdquo International Journal ofAutomotive Technology vol 20 no 5 pp 1051ndash1063 2019

[4] H Farah and H N Koutsopoulos ldquoDo cooperative systemsmake driversrsquo car-following behavior saferrdquo TransportationResearch Part C Emerging Technologies vol 41 pp 61ndash722014

[5] SAE Taxonomy and Definitions for Terms Related to DrivingAutomation Systems for On-Road Motor Vehicles SAEJ3016TM Warrendale PA USA 2018

[6] J Hoegg J W Alba and D W Dahl ldquoe good the bad andthe ugly influence of aesthetics on product feature judg-mentsrdquo Journal of Consumer Psychology vol 20 no 4pp 419ndash430 2010

[7] M Kyriakidis R Happee and J C F De Winter ldquoPublicopinion on automated driving results of an internationalquestionnaire among 5000 respondentsrdquo TransportationResearch Part F Traffic Psychology and Behaviour vol 32pp 127ndash140 2015

[8] F Meng and C Spence ldquoTactile warning signals for in-vehiclesystemsrdquo Accident Analysis amp Prevention vol 75 pp 333ndash346 2015

[9] J Wan and C Wu ldquoe effects of lead time of take-overrequest and nondriving tasks on taking-over control of au-tomated vehiclesrdquo IEEE Transactions on Human MachineSystems vol 48 no 6 pp 582ndash591 2018

[10] S Petermeijer P Hornberger I Ganotis J D Winter andK Bengler ldquoe design of a vibrotactile seat for conveyingtake-over requests in automated drivingrdquo in Proceedings of the8th International Conference on Applied Human Factors andErgonomics Los Angeles CA USA July 2017

[11] T Ito A Takata and K Oosawa ldquoTime required for take-overfrom automated to manual drivingrdquo in Proceedings of the SAE2016 World Congress and Exhibition Detroit MI USA April2016

[12] K Zeeb A Buchner and M Schrauf ldquoWhat determines thetake-over time An integrated model approach of driver take-over after automated drivingrdquo Accident Analysis amp Preven-tion vol 78 pp 212ndash221 2015

[13] B Mok M Johns K J Lee et al ldquoEmergency automation offunstructured transition timing for distracted drivers of au-tomated vehiclesrdquo in Proceedings of the IEEE InternationalConference on Intelligent Transportation Systems Las PalmasSpain September 2015

[14] S Samuel A Borowsky S Zilberstein and D L FisherldquoMinimum time to situation awareness in scenarios involvingtransfer of control from an automated driving suiterdquoTransportation Research Record Journal of the TransportationResearch Board vol 2602 no 2602 pp 115ndash120 2016

[15] W Vlakveld N van Nes J de Bruin L Vissers andM van der Kroft ldquoSituation awareness increases when drivershave more time to take over the wheel in a level 3 automated

Complexity 11

car a simulator studyrdquo Transportation Research Part F TrafficPsychology and Behaviour vol 58 pp 917ndash929 2018

[16] N Merat A H Jamson F C H Lai M Daly andO M J Carsten ldquoTransition to manual driver behaviourwhen resuming control from a highly automated vehiclerdquoTransportation Research Part F Traffic Psychology and Be-haviour vol 27 no 26 pp 274ndash282 2014

[17] C Z Wu H R Wu and N C Lyu ldquoReview of control switchand safety of human-computer driving intelligent vehiclerdquoJournal of Traffic and Transportation Engineering vol 18no 6 pp 131ndash141 2018

[18] C Gold D Dambock L Lorenz and K Bengler ldquoldquoTakeoverrdquo how long does it take to get the driver back into thelooprdquo Proceedings of the Human Factors and ErgonomicsSociety Annual Meeting vol 57 no 1 pp 1938ndash1942 2013

[19] H Yang Y Zhao and Y Wang ldquoIdentifying modeling formsof instrument panel system in intelligent shared cars a studyfor perceptual preference and in-vehicle behaviorsrdquo Envi-ronmental Science and Pollution Research vol 27 no 1pp 1009ndash1023 2020

[20] A Berkovich A Lu B Levine and A V Reddy ldquoObservedcustomer seating and standing behavior and seat preferenceson board subway cars in New York cityrdquo TransportationResearch Record Journal of the Transportation ResearchBoard vol 2353 no 1 pp 33ndash46 2013

[21] R Li Y V Chen C Sha and Z Lu ldquoEffects of interface layouton the usability of in-vehicle information systems and drivingsafetyrdquo Displays vol 49 pp 124ndash132 2017

[22] H Kim S Kwon J Heo H Lee and M K Chung ldquoe effectof touch-key size on the usability of in-vehicle informationsystems and driving safety during simulated drivingrdquo AppliedErgonomics vol 45 no 3 pp 379ndash388 2014

[23] H Kim and H Song ldquoEvaluation of the safety and usability oftouch gestures in operating in-vehicle information systemswith visual occlusionrdquo Applied Ergonomics vol 45 no 3pp 789ndash798 2014

[24] M Koerber C Gold D Lechner and K Bengler ldquoe in-fluence of age on the take-over of vehicle control in highlyautomated drivingrdquo Transportation Research Part F TrafficPsychology amp Behaviour vol 39 pp 19ndash32 2016

[25] H Clark A C Mclaughlin B Williams and J Feng ldquoPer-formance in takeover and characteristics of non-driving re-lated tasks during highly automated driving in younger andolder driversrdquo Proceedings of the Human Factors amp Ergo-nomics Society Annual Meeting vol 61 no 1 pp 37ndash41 2017

[26] I Lijarcio S A Useche J Llamazares and L MontoroldquoPerceived benefits and constraints in vehicle automationdata to assess the relationship between driverrsquos features andtheir attitudes towards autonomous vehiclesrdquo Data in Briefvol 27 Article ID 104662 2019

[27] K-Z Chen and X-A Feng ldquoVirtual genes of manufacturingproducts and their reforms for product innovative designrdquoProceedings of the Institution of Mechanical Engineers Part CJournal of Mechanical Engineering Science vol 218 no 5pp 557ndash574 2004

[28] K-Z Chen X-A Feng and X-C Chen ldquoReverse deductionof virtual chromosomes of manufactured products for theirgene-engineering-based innovative designrdquo Computer-AidedDesign vol 37 no 11 pp 1191ndash1203 2005

[29] H-C Tsai and J-R Chou ldquoAutomatic design support andimage evaluation of two-coloured products using colour as-sociation and colour harmony scales and genetic algorithmrdquoComputer-Aided Design vol 39 no 9 pp 818ndash828 2007

[30] S-W Hsiao C-F Hsu and K-W Tang ldquoA consultation andsimulation system for product color planning based on in-teractive genetic algorithmsrdquo Color Research amp Applicationvol 38 no 5 pp 375ndash390 2013

[31] Y Sun W-l Wang Y-w Zhao and X-j Liu ldquoUser imageoriented interactive genetic algorithm evaluation mode inproduct developmentrdquo Computer Integrated ManufacturingSystems vol 18 no 2 pp 276ndash281 2012

[32] R Kamalian E Yeh Z Ying A M Agogino and H TakagildquoReducing human fatigue in interactive evolutionary com-putation through fuzzy systems and machine learning sys-temsrdquo in Proceedings of the 2006 IEEE InternationalConference on Fuzzy Systems Vancouver Canada July 2006

[33] Z Gu M Xi Tang and J H Frazer ldquoCapturing aestheticintention during interactive evolutionrdquo Computer-AidedDesign vol 38 no 3 pp 224ndash237 2006

[34] G Harih M Borovinsek and Z Ren Optimisation ofProductrsquos Hand-Handle Interface Material Parameters forImproved Ergonomics Springer International PublishingCham Switzerland 2015

[35] P Cheng D Chen and J Wang ldquoClustering of the bodyshape of the adult male by using principal component analysisand genetic algorithm-BP neural networkrdquo Soft Computingvol 24 no 17 Article ID 13219 2020

[36] N M B Serrano P J G Nieto A S Sanchez F S Lasherasand P R Fernandez AHybrid Algorithm for the Assessment ofthe Influence of Risk Factors in the Development of Upper LimbMusculoskeletal Disorders Springer Cham Switzerland 2018

[37] S S Sankar O-M Holman A F Gazabon and C J AcevedoldquoApplication of genetic algorithm to job scheduling underergonomic constraints in manufacturing industryrdquo Journal ofAmbient Intelligence and Humanized Computing vol 10no 5 pp 2063ndash2090 2019

[38] A-Z Atiya L Lee and K Xing ldquoDeveloping a multi-ob-jective genetic optimisation approach for an operationaldesign of a manual mixed-model assembly line with walkingworkersrdquo Journal of Intelligent Manufacturing vol 27 no 5pp 1ndash17 2014

[39] P Jedlicka and T Ryba ldquoGenetic algorithm application inimage segmentationrdquo Pattern Recognition and Image Anal-ysis vol 26 no 3 pp 497ndash501 2016

[40] L V Campenhout J Frens K Overbeeke A Standaert andH Peremans ldquoPhysical interaction in a dematerializedworldrdquo International Journal of Design vol 7 no 1 pp 1ndash182013

[41] K Nandhini and S R Balasundaram ldquoImproving readabilitythrough individualized summary extraction using interactivegenetic algorithmrdquo Applied Artificial Intelligence vol 30no 7 pp 635ndash661 2016

[42] R M Gray Entropy and Information 6eory Springer NewYork NY USA 2nd edition 2011

[43] B Anderhofstadt and S Spinler ldquoFactors affecting the pur-chasing decision and operation of alternative fuel-poweredheavy-duty trucks in Germanymdasha Delphi studyrdquo Trans-portation Research Part D Transport and Environmentvol 73 pp 87ndash107 2019

[44] X-m Wu and Y Feng ldquoReactive power optimization ofdistribution network based on improved genetic algorithmrdquoJournal of Xirsquoan Shiyou University (Natural Science Edition)vol 30 no 3 pp 95ndash99 2015

[45] Z Ling andM-j Li ldquoState space evolutionary algorithm basedon non-uniform mutation operatorrdquo Computer Technologyand Development vol 28 no 9 pp 68ndash71 2018

12 Complexity

[46] W Hu and J Zhao ldquoAutomobile styling gene evolutiondriven by usersrsquo expectation imagerdquo Journal of MechanicalEngineering vol 47 no 16 pp 176ndash181 2011

[47] J J Scott and R Gray ldquoA comparison of tactile visual andauditory warnings for rear-end collision prevention in sim-ulated drivingrdquo Human Factors 6e Journal of the HumanFactors and Ergonomics Society vol 50 no 2 pp 264ndash2752008

[48] B Reimer B Mehler and J F Coughlin ldquoReductions in self-reported stress and anticipatory heart rate with the use of asemi-automated parallel parking systemrdquo Applied Ergonom-ics vol 52 pp 120ndash127 2016

[49] B DrsquoIppolito ldquoe importance of design for firms com-petitiveness a review of the literaturerdquo Technovation vol 34no 11 pp 716ndash730 2014

[50] S Asensio-Cuesta J A Diego-Mas L Canos-Daros andC Andres-Romano ldquoA genetic algorithm for the design of jobrotation schedules considering ergonomic and competencecriteriardquo International Journal of Advanced ManufacturingTechnology vol 60 no 9ndash12 pp 1161ndash1174 2012

[51] S-C Lee H-E Tseng C-C Chang and Y-M HuangldquoApplying interactive genetic algorithms to disassembly se-quence planningrdquo International Journal of Precision Engi-neering and Manufacturing vol 21 no 4 pp 663ndash679 2020

[52] M G Shishaev V V Dikovitsky and L V Lapochkina ldquoeexperience of building cognitive user interfaces of multido-main information systems based on the mental model ofusersrdquo in Proceedings of the 6th Computer Science On-LineConference (CSOC 2017) Prague Czech Republic April 2017

[53] S Lee H Alzoubi and S Kim ldquoe effect of interior designelements and lighting layouts on prospective occupantsrsquoperceptions of amenity and efficiency in living roomsrdquo Sus-tainability vol 9 no 7 Article ID 1119 2017

[54] G Caruso ldquoMixed reality system for ergonomic assessment ofdriverrsquos seatrdquo International Journal of Virtual Reality vol 10no 2 pp 69ndash79 2011

[55] M L Reyes and J D Lee ldquoEffects of cognitive load presenceand duration on driver eye movements and event detectionperformancerdquo Transportation Research Part F Traffic Psy-chology and Behaviour vol 11 no 6 pp 391ndash402 2008

[56] D Cisler P M Greenwood D M Roberts R McKendrickand C L Baldwin ldquoComparing the relative strengths of EEGand low-cost physiological devices in modeling attentionallocation in semiautonomous vehiclesrdquo Frontiers in HumanNeuroscience vol 13 Article ID 109 2019

[57] I-H Kim J-W Kim S Haufe and S-W Lee ldquoDetection ofbraking intention in diverse situations during simulateddriving based on EEG feature combinationrdquo Journal of NeuralEngineering vol 12 no 1 Article ID 016001 2014

Complexity 13

adaptive cruise the completion time of operating with a sidestalk was shorter (B) for the task of adjusting the distancebetween cars the completion time would be shorter whenoperating with a button located in the middle of the steeringwheel spoke (C) for the task of auto hold a starting buttonon the left side of the EPB button could bring about a shortercompletion time ese results provided a reference for thefuture automobile interior design

5 Discussion

Driving assistant system (DAS) plays an important role inthe context of human-computer codriving As reportedDAS can help the driver to complete the intense drivingtasks effectively and abate the pressure level of the driver[48] erefore it is meaningful to make a study for thecontroller dimensions in the DAS e study collectedcomfortable dimensions perceived by the drivers in a relaxedstate e aim was to detect a set of reasonable dimensions

which could make drivers at a low level of stress In ad-dition it is valuable for drivers to adapt to the different roadenvironment and realize sustainable traffic to optimize thesizes of interactive components on the in-vehicle HMI emain goals of in-vehicle technologies and cooperativeservices were to reduce traffic congestion and improve roadsafety [4]

e study attempted to analyze the optimal HMI di-mensions suitable for take-over tasks e involved HMIcomponents such as the steering wheel self-adaptive cruiselever distance adjustment button and auto hold buttonwere all direct design elements for users to see and touchBecause the design and performance of a product wereimportant factors affecting the decision-making of con-sumers [49] we calculated the information entropy andinformation gains of the three kinds of take-over tasks forclassifying usersrsquo purchase intentions e informationtransmitted by a product was multivariant complex andfuzzy so that there would be dissipation in the process oftransmission erefore the image evaluation process ofproducts could be regarded as a chaotic state e feedbackinformation could be calculated by information entropy tocarry out a comprehensive evaluation e smaller the en-tropy of an image the clearer it was From the perspective ofchaos degree the 2nd factor in Table 3 (the position of thefunction key to adjust the distance between cars) broughtmore uncertainty to the purchase decision(E(A2) 09592) which indicated that little informationwas provided by the 2nd factor for judging the usersrsquo de-cisions On the contrary the uncertainty produced by the 1stfactor (the starting mode of the self-adaptive cruise) was thesmallest (E(A1) 07842) is illustrated that using theside stalk or a button to complete the switch of the controlright had a great impact on the satisfaction and purchasedecision of users

After normalizing the information entropy of threekinds of take-over tasks their weight values were calculated

Table 5 e finally generated dimensional data of the individuals with the highest fitness

(A) e starting mode of self-adaptive cruise

(B) e position of thefunction key to adjust thedistance between cars

(C) e position of the function key of auto-hold Fitness value

a11 a12 a21 a22 b11 b12 b21 b22 c11 c12 c21 c22 c31 c32202 062 089 009 136 004 148 030 284 302 271 338 321 313 008

Table 6 e paired sample t-test results of task completion time under the two HMIs with different dimensions

Interaction modes of take-over tasks

e task completiontime under the original

dimensions (s)

e task completiontime under the evolved

dimensions (s) t-value Sig

Mean Std dev Mean Std devA1lowast 0715 0185 0537 0167 2553 0025A2 0657 0181 0637 0186 1151 0272B1lowast 0648 013 0462 0206 2797 0016B2lowastlowast 0736 0106 0571 0124 3116 0009C1lowast 0753 0295 0562 0222 2515 0027C2 0667 0226 0522 0176 1999 0069C3lowast 069 0208 0537 0225 2355 0036

A1lowast B1lowast B2lowastlowast C1lowast C3lowast

0

02

04

06

08

1

12

Time under the original dimensionsTime under the evolved dimensions

Figure 4 e completion time of the tasks with significant dif-ferences under the two HMIs with different dimensions

Complexity 9

and the fitness function was constructed In this way thedimensions with the highest fitness could be found out bymeans of GA

e background and driving experience of the subjects weredifferent and their familiarity with the take-over tasks and theHMI was variant Besides the noise of usersrsquo subjective per-ception data was always high Taking the above factors intoconsideration we collected PCD data repeatedly from the same13 subjects to form the initial population From the evolutionarytrajectory it could be found that in the 8th 10th 11th 14th 17thand 21st iteration there appeared an optimal solution

GA was always used for parameter optimization bysearching the optimal solutions while seldom used in userresearch and environment perception analysis But in studiesof user-centred design it could search the set of satisfactorysolutions driven by user expectations [46] Many pieces ofresearch had utilized GA to improve the ergonomic effi-ciency or optimize interactive component sizes in an op-erating environment [34ndash38 50 51] In this study wereported a driver perception research which investigated themost comfortable HMI sizes under different interactionmodes of take-over tasks By means of normalized geometricselection arithmetic uniform crossover and nonuniformmutation GA was applied in the research of perceivedcomfortable dimensions and a set of reasonable dimensionswas obtained e dimensional evolutionary method dis-cussed in this study embodied the concept of dynamic in-teraction design which took the mutual influence andrestrictions among different dimensions into considerationCompared with traditional methods of measuring manytimes and taking the average this method was more ad-vanced And the study enriched theoretical researches in thefield of product design and collaborative optimization

Perceived comfort was a complex problem for its dif-ficulty to be quantified objectively But it was useful in er-gonomic explorations Researchers had found that theperceptual messages from product property and usage re-quirements were competent to the analysis of user experi-ence [52] erefore it was reasonable to probe suitabledimensions by perceived comfort For the study of HMIdesign effect in a specific environment these data couldproduce important and dependable results [53] Rationaldimensions of the automobile interior HMI played an im-portant role in the driverrsquos operation quality Caruso [54]studied driversrsquo experience on the seat by an experiment andfound that interior space influenced the driversrsquo ergonomicperformance [54] For the take-over tasks in the situation ofhuman-computer codriving the dimensions also affectedwhether the driver could safely achieve the control rightswitch in the shortest possible time

In this paper paired-sample t-test was used to testwhether the sizes with the highest fitness could bring about asignificant reduction in task completion time In the resultswe found that when the interface adopted the finally evolvedsizes the completion time of each task was shorter As thecompletion time of in-vehicle interactive tasks was related tothe vehicle speed and driver errors under the context ofhuman-computer codriving [21 55] a shorter completiontime would help the driver to control the speed better

6 Limitation

e study explored the reasonable dimensions of the hu-man-machine interface in intelligent cars Neverthelessthere were still some deficiencies in the equipment used inthis study e driving simulator was easier to control andreform and helped to reduce the subjectsrsquo pressure How-ever the immersion visual interactivity and fidelity degreewere not high enough After obtaining reliable conclusionsin this study outdoor experiment with a real vehicle could beconducted in the next stage

Besides the sample size of this study was limited Whenthe sample size was larger in future studies the represen-tativeness of the data would be better and the results of GAcould be further improved

Finally usersrsquo perception of the HMI dimensions de-viated from each other In the next phase of the studyelectroencephalography (EEG) signal in the process ofperforming the driving tasks could be used By this meansthe subjectsrsquo attention and intention when interacting withthe HMI could be reflected in an objective way [56 57] as asupplement to their subjective report

7 Conclusion

For HMI design of industrial products physical concretemodelling parameters such as the product shape size andlayout were the objective properties that users could directlyperceive and the basis for the product to be expressed Basedon the dynamic measurement principle this paper proposedan optimized design method for ergonomic dimensions byGA e results indicated that the method was effective andprovided a new way for product shape innovation andhuman-machine dimensional automatic design e mainconclusions of this study were as follows

(1) Based on the information entropy theory the weightvalues of the three driving assistance functions werecalculated e results showed that the informationgain of the starting mode of the self-adaptive cruisewas highest while the information gain of the po-sition of the function key to adjust the distancebetween cars was lowest which provided a theo-retical basis for the dimensional fitness evaluationmechanism of the HMI

(2) e evolutionary process of HMI dimensionsdriven by user expectation was studied and theevolution conditions were analyzed e studyconfirmed the feasibility of normalized geometricselection arithmetic uniform crossover and non-uniform mutation in iterative researches for hu-man-machine dimensions rough parameterdebugging this evolutionary method was proved tobe effective in obtaining appropriate sizes of HMIcomponents e model could make full use of theadvantages of GA to improve the user researchperformance such as few restrictions on the opti-mization high robustness and an easily achievedglobal search

10 Complexity

(3) After finding the dimensions with the highest fitnessand applying them to the design of the HMI within-subjects experiments showed that the completiontime of five take-over tasks was significantly shorterexcept for that of Task A2 and C2is indicated thatthe information processing of drivers under the HMIwith evolved dimensions was faster which was morebeneficial to the vehicle control and take-overquality Intelligent vehicles achieved intelligent in-formation exchange with people vehicles and roadsthrough in-vehicle sensor system and informationterminals erefore the evolved dimensions couldmake the vehicle apt to respond to the environmentand have more time to fully deal with the complextraffic environment and would help to improve thetraffic efficiency therefrom

Data Availability

e data used to support the findings of this study areavailable from the corresponding author upon request

Consent

e study respected and guaranteed the subjectsrsquo right toparticipate in the study and allowed them to withdraw from thestudy unconditionally at any stage e study strictly followedthe informed consent procedure e subjects were asked tosign a consent form to inform them of the purpose processduration of the experiment matters needing attention and themethods and purpose of collecting the data so as to reducetheir anxiety and confusion In the experiment the personalsafety and health of the subjects took priority According to theusage guidelines of the driving simulator we tried to avoid anypossible injury in driving tasks e privacy of the subjects wasstrictly protected e subjects were informed of the storageuse and confidentiality measures of their personal informationand image data in advance truthfully And we would notdisclose the information to a third party or spread them on theInternet without authorization

Conflicts of Interest

e authors declare that they have no conflicts of interest

Acknowledgments

is researchwas funded by BeijingUrbanGovernance ResearchProject (grant no 20XN244) Scientific Research Program ofBeijing Education Commission (grant no KM202010009003)Yuyou Talent Support Program of North China University ofTechnology (grant no 107051360018XN012018) Tianjin YouthTalents Reserve Support Program and Chinese ErgonomicsSociety amp Kingfar Joint Research Fund for Outstanding YoungScholars (grant number CES-Kingfar-2019-001)

References

[1] S Jin J Yang E Wang and J Liu ldquoe influence of high-speed rail on ice-snow tourism in north eastern ChinardquoTourism Management vol 78 Article ID 104070 2020

[2] J Yang R Yang J Sun T Huang and Q Ge ldquoe spatialdifferentiation of the suitability of ice-snow tourist destina-tions based on a comprehensive evaluation model in ChinardquoSustainability vol 9 no 5 Article ID 774 2017

[3] W-Y Chung T-W Chong and B-G Lee ldquoMethods to detectand reduce driver stress a reviewrdquo International Journal ofAutomotive Technology vol 20 no 5 pp 1051ndash1063 2019

[4] H Farah and H N Koutsopoulos ldquoDo cooperative systemsmake driversrsquo car-following behavior saferrdquo TransportationResearch Part C Emerging Technologies vol 41 pp 61ndash722014

[5] SAE Taxonomy and Definitions for Terms Related to DrivingAutomation Systems for On-Road Motor Vehicles SAEJ3016TM Warrendale PA USA 2018

[6] J Hoegg J W Alba and D W Dahl ldquoe good the bad andthe ugly influence of aesthetics on product feature judg-mentsrdquo Journal of Consumer Psychology vol 20 no 4pp 419ndash430 2010

[7] M Kyriakidis R Happee and J C F De Winter ldquoPublicopinion on automated driving results of an internationalquestionnaire among 5000 respondentsrdquo TransportationResearch Part F Traffic Psychology and Behaviour vol 32pp 127ndash140 2015

[8] F Meng and C Spence ldquoTactile warning signals for in-vehiclesystemsrdquo Accident Analysis amp Prevention vol 75 pp 333ndash346 2015

[9] J Wan and C Wu ldquoe effects of lead time of take-overrequest and nondriving tasks on taking-over control of au-tomated vehiclesrdquo IEEE Transactions on Human MachineSystems vol 48 no 6 pp 582ndash591 2018

[10] S Petermeijer P Hornberger I Ganotis J D Winter andK Bengler ldquoe design of a vibrotactile seat for conveyingtake-over requests in automated drivingrdquo in Proceedings of the8th International Conference on Applied Human Factors andErgonomics Los Angeles CA USA July 2017

[11] T Ito A Takata and K Oosawa ldquoTime required for take-overfrom automated to manual drivingrdquo in Proceedings of the SAE2016 World Congress and Exhibition Detroit MI USA April2016

[12] K Zeeb A Buchner and M Schrauf ldquoWhat determines thetake-over time An integrated model approach of driver take-over after automated drivingrdquo Accident Analysis amp Preven-tion vol 78 pp 212ndash221 2015

[13] B Mok M Johns K J Lee et al ldquoEmergency automation offunstructured transition timing for distracted drivers of au-tomated vehiclesrdquo in Proceedings of the IEEE InternationalConference on Intelligent Transportation Systems Las PalmasSpain September 2015

[14] S Samuel A Borowsky S Zilberstein and D L FisherldquoMinimum time to situation awareness in scenarios involvingtransfer of control from an automated driving suiterdquoTransportation Research Record Journal of the TransportationResearch Board vol 2602 no 2602 pp 115ndash120 2016

[15] W Vlakveld N van Nes J de Bruin L Vissers andM van der Kroft ldquoSituation awareness increases when drivershave more time to take over the wheel in a level 3 automated

Complexity 11

car a simulator studyrdquo Transportation Research Part F TrafficPsychology and Behaviour vol 58 pp 917ndash929 2018

[16] N Merat A H Jamson F C H Lai M Daly andO M J Carsten ldquoTransition to manual driver behaviourwhen resuming control from a highly automated vehiclerdquoTransportation Research Part F Traffic Psychology and Be-haviour vol 27 no 26 pp 274ndash282 2014

[17] C Z Wu H R Wu and N C Lyu ldquoReview of control switchand safety of human-computer driving intelligent vehiclerdquoJournal of Traffic and Transportation Engineering vol 18no 6 pp 131ndash141 2018

[18] C Gold D Dambock L Lorenz and K Bengler ldquoldquoTakeoverrdquo how long does it take to get the driver back into thelooprdquo Proceedings of the Human Factors and ErgonomicsSociety Annual Meeting vol 57 no 1 pp 1938ndash1942 2013

[19] H Yang Y Zhao and Y Wang ldquoIdentifying modeling formsof instrument panel system in intelligent shared cars a studyfor perceptual preference and in-vehicle behaviorsrdquo Envi-ronmental Science and Pollution Research vol 27 no 1pp 1009ndash1023 2020

[20] A Berkovich A Lu B Levine and A V Reddy ldquoObservedcustomer seating and standing behavior and seat preferenceson board subway cars in New York cityrdquo TransportationResearch Record Journal of the Transportation ResearchBoard vol 2353 no 1 pp 33ndash46 2013

[21] R Li Y V Chen C Sha and Z Lu ldquoEffects of interface layouton the usability of in-vehicle information systems and drivingsafetyrdquo Displays vol 49 pp 124ndash132 2017

[22] H Kim S Kwon J Heo H Lee and M K Chung ldquoe effectof touch-key size on the usability of in-vehicle informationsystems and driving safety during simulated drivingrdquo AppliedErgonomics vol 45 no 3 pp 379ndash388 2014

[23] H Kim and H Song ldquoEvaluation of the safety and usability oftouch gestures in operating in-vehicle information systemswith visual occlusionrdquo Applied Ergonomics vol 45 no 3pp 789ndash798 2014

[24] M Koerber C Gold D Lechner and K Bengler ldquoe in-fluence of age on the take-over of vehicle control in highlyautomated drivingrdquo Transportation Research Part F TrafficPsychology amp Behaviour vol 39 pp 19ndash32 2016

[25] H Clark A C Mclaughlin B Williams and J Feng ldquoPer-formance in takeover and characteristics of non-driving re-lated tasks during highly automated driving in younger andolder driversrdquo Proceedings of the Human Factors amp Ergo-nomics Society Annual Meeting vol 61 no 1 pp 37ndash41 2017

[26] I Lijarcio S A Useche J Llamazares and L MontoroldquoPerceived benefits and constraints in vehicle automationdata to assess the relationship between driverrsquos features andtheir attitudes towards autonomous vehiclesrdquo Data in Briefvol 27 Article ID 104662 2019

[27] K-Z Chen and X-A Feng ldquoVirtual genes of manufacturingproducts and their reforms for product innovative designrdquoProceedings of the Institution of Mechanical Engineers Part CJournal of Mechanical Engineering Science vol 218 no 5pp 557ndash574 2004

[28] K-Z Chen X-A Feng and X-C Chen ldquoReverse deductionof virtual chromosomes of manufactured products for theirgene-engineering-based innovative designrdquo Computer-AidedDesign vol 37 no 11 pp 1191ndash1203 2005

[29] H-C Tsai and J-R Chou ldquoAutomatic design support andimage evaluation of two-coloured products using colour as-sociation and colour harmony scales and genetic algorithmrdquoComputer-Aided Design vol 39 no 9 pp 818ndash828 2007

[30] S-W Hsiao C-F Hsu and K-W Tang ldquoA consultation andsimulation system for product color planning based on in-teractive genetic algorithmsrdquo Color Research amp Applicationvol 38 no 5 pp 375ndash390 2013

[31] Y Sun W-l Wang Y-w Zhao and X-j Liu ldquoUser imageoriented interactive genetic algorithm evaluation mode inproduct developmentrdquo Computer Integrated ManufacturingSystems vol 18 no 2 pp 276ndash281 2012

[32] R Kamalian E Yeh Z Ying A M Agogino and H TakagildquoReducing human fatigue in interactive evolutionary com-putation through fuzzy systems and machine learning sys-temsrdquo in Proceedings of the 2006 IEEE InternationalConference on Fuzzy Systems Vancouver Canada July 2006

[33] Z Gu M Xi Tang and J H Frazer ldquoCapturing aestheticintention during interactive evolutionrdquo Computer-AidedDesign vol 38 no 3 pp 224ndash237 2006

[34] G Harih M Borovinsek and Z Ren Optimisation ofProductrsquos Hand-Handle Interface Material Parameters forImproved Ergonomics Springer International PublishingCham Switzerland 2015

[35] P Cheng D Chen and J Wang ldquoClustering of the bodyshape of the adult male by using principal component analysisand genetic algorithm-BP neural networkrdquo Soft Computingvol 24 no 17 Article ID 13219 2020

[36] N M B Serrano P J G Nieto A S Sanchez F S Lasherasand P R Fernandez AHybrid Algorithm for the Assessment ofthe Influence of Risk Factors in the Development of Upper LimbMusculoskeletal Disorders Springer Cham Switzerland 2018

[37] S S Sankar O-M Holman A F Gazabon and C J AcevedoldquoApplication of genetic algorithm to job scheduling underergonomic constraints in manufacturing industryrdquo Journal ofAmbient Intelligence and Humanized Computing vol 10no 5 pp 2063ndash2090 2019

[38] A-Z Atiya L Lee and K Xing ldquoDeveloping a multi-ob-jective genetic optimisation approach for an operationaldesign of a manual mixed-model assembly line with walkingworkersrdquo Journal of Intelligent Manufacturing vol 27 no 5pp 1ndash17 2014

[39] P Jedlicka and T Ryba ldquoGenetic algorithm application inimage segmentationrdquo Pattern Recognition and Image Anal-ysis vol 26 no 3 pp 497ndash501 2016

[40] L V Campenhout J Frens K Overbeeke A Standaert andH Peremans ldquoPhysical interaction in a dematerializedworldrdquo International Journal of Design vol 7 no 1 pp 1ndash182013

[41] K Nandhini and S R Balasundaram ldquoImproving readabilitythrough individualized summary extraction using interactivegenetic algorithmrdquo Applied Artificial Intelligence vol 30no 7 pp 635ndash661 2016

[42] R M Gray Entropy and Information 6eory Springer NewYork NY USA 2nd edition 2011

[43] B Anderhofstadt and S Spinler ldquoFactors affecting the pur-chasing decision and operation of alternative fuel-poweredheavy-duty trucks in Germanymdasha Delphi studyrdquo Trans-portation Research Part D Transport and Environmentvol 73 pp 87ndash107 2019

[44] X-m Wu and Y Feng ldquoReactive power optimization ofdistribution network based on improved genetic algorithmrdquoJournal of Xirsquoan Shiyou University (Natural Science Edition)vol 30 no 3 pp 95ndash99 2015

[45] Z Ling andM-j Li ldquoState space evolutionary algorithm basedon non-uniform mutation operatorrdquo Computer Technologyand Development vol 28 no 9 pp 68ndash71 2018

12 Complexity

[46] W Hu and J Zhao ldquoAutomobile styling gene evolutiondriven by usersrsquo expectation imagerdquo Journal of MechanicalEngineering vol 47 no 16 pp 176ndash181 2011

[47] J J Scott and R Gray ldquoA comparison of tactile visual andauditory warnings for rear-end collision prevention in sim-ulated drivingrdquo Human Factors 6e Journal of the HumanFactors and Ergonomics Society vol 50 no 2 pp 264ndash2752008

[48] B Reimer B Mehler and J F Coughlin ldquoReductions in self-reported stress and anticipatory heart rate with the use of asemi-automated parallel parking systemrdquo Applied Ergonom-ics vol 52 pp 120ndash127 2016

[49] B DrsquoIppolito ldquoe importance of design for firms com-petitiveness a review of the literaturerdquo Technovation vol 34no 11 pp 716ndash730 2014

[50] S Asensio-Cuesta J A Diego-Mas L Canos-Daros andC Andres-Romano ldquoA genetic algorithm for the design of jobrotation schedules considering ergonomic and competencecriteriardquo International Journal of Advanced ManufacturingTechnology vol 60 no 9ndash12 pp 1161ndash1174 2012

[51] S-C Lee H-E Tseng C-C Chang and Y-M HuangldquoApplying interactive genetic algorithms to disassembly se-quence planningrdquo International Journal of Precision Engi-neering and Manufacturing vol 21 no 4 pp 663ndash679 2020

[52] M G Shishaev V V Dikovitsky and L V Lapochkina ldquoeexperience of building cognitive user interfaces of multido-main information systems based on the mental model ofusersrdquo in Proceedings of the 6th Computer Science On-LineConference (CSOC 2017) Prague Czech Republic April 2017

[53] S Lee H Alzoubi and S Kim ldquoe effect of interior designelements and lighting layouts on prospective occupantsrsquoperceptions of amenity and efficiency in living roomsrdquo Sus-tainability vol 9 no 7 Article ID 1119 2017

[54] G Caruso ldquoMixed reality system for ergonomic assessment ofdriverrsquos seatrdquo International Journal of Virtual Reality vol 10no 2 pp 69ndash79 2011

[55] M L Reyes and J D Lee ldquoEffects of cognitive load presenceand duration on driver eye movements and event detectionperformancerdquo Transportation Research Part F Traffic Psy-chology and Behaviour vol 11 no 6 pp 391ndash402 2008

[56] D Cisler P M Greenwood D M Roberts R McKendrickand C L Baldwin ldquoComparing the relative strengths of EEGand low-cost physiological devices in modeling attentionallocation in semiautonomous vehiclesrdquo Frontiers in HumanNeuroscience vol 13 Article ID 109 2019

[57] I-H Kim J-W Kim S Haufe and S-W Lee ldquoDetection ofbraking intention in diverse situations during simulateddriving based on EEG feature combinationrdquo Journal of NeuralEngineering vol 12 no 1 Article ID 016001 2014

Complexity 13

and the fitness function was constructed In this way thedimensions with the highest fitness could be found out bymeans of GA

e background and driving experience of the subjects weredifferent and their familiarity with the take-over tasks and theHMI was variant Besides the noise of usersrsquo subjective per-ception data was always high Taking the above factors intoconsideration we collected PCD data repeatedly from the same13 subjects to form the initial population From the evolutionarytrajectory it could be found that in the 8th 10th 11th 14th 17thand 21st iteration there appeared an optimal solution

GA was always used for parameter optimization bysearching the optimal solutions while seldom used in userresearch and environment perception analysis But in studiesof user-centred design it could search the set of satisfactorysolutions driven by user expectations [46] Many pieces ofresearch had utilized GA to improve the ergonomic effi-ciency or optimize interactive component sizes in an op-erating environment [34ndash38 50 51] In this study wereported a driver perception research which investigated themost comfortable HMI sizes under different interactionmodes of take-over tasks By means of normalized geometricselection arithmetic uniform crossover and nonuniformmutation GA was applied in the research of perceivedcomfortable dimensions and a set of reasonable dimensionswas obtained e dimensional evolutionary method dis-cussed in this study embodied the concept of dynamic in-teraction design which took the mutual influence andrestrictions among different dimensions into considerationCompared with traditional methods of measuring manytimes and taking the average this method was more ad-vanced And the study enriched theoretical researches in thefield of product design and collaborative optimization

Perceived comfort was a complex problem for its dif-ficulty to be quantified objectively But it was useful in er-gonomic explorations Researchers had found that theperceptual messages from product property and usage re-quirements were competent to the analysis of user experi-ence [52] erefore it was reasonable to probe suitabledimensions by perceived comfort For the study of HMIdesign effect in a specific environment these data couldproduce important and dependable results [53] Rationaldimensions of the automobile interior HMI played an im-portant role in the driverrsquos operation quality Caruso [54]studied driversrsquo experience on the seat by an experiment andfound that interior space influenced the driversrsquo ergonomicperformance [54] For the take-over tasks in the situation ofhuman-computer codriving the dimensions also affectedwhether the driver could safely achieve the control rightswitch in the shortest possible time

In this paper paired-sample t-test was used to testwhether the sizes with the highest fitness could bring about asignificant reduction in task completion time In the resultswe found that when the interface adopted the finally evolvedsizes the completion time of each task was shorter As thecompletion time of in-vehicle interactive tasks was related tothe vehicle speed and driver errors under the context ofhuman-computer codriving [21 55] a shorter completiontime would help the driver to control the speed better

6 Limitation

e study explored the reasonable dimensions of the hu-man-machine interface in intelligent cars Neverthelessthere were still some deficiencies in the equipment used inthis study e driving simulator was easier to control andreform and helped to reduce the subjectsrsquo pressure How-ever the immersion visual interactivity and fidelity degreewere not high enough After obtaining reliable conclusionsin this study outdoor experiment with a real vehicle could beconducted in the next stage

Besides the sample size of this study was limited Whenthe sample size was larger in future studies the represen-tativeness of the data would be better and the results of GAcould be further improved

Finally usersrsquo perception of the HMI dimensions de-viated from each other In the next phase of the studyelectroencephalography (EEG) signal in the process ofperforming the driving tasks could be used By this meansthe subjectsrsquo attention and intention when interacting withthe HMI could be reflected in an objective way [56 57] as asupplement to their subjective report

7 Conclusion

For HMI design of industrial products physical concretemodelling parameters such as the product shape size andlayout were the objective properties that users could directlyperceive and the basis for the product to be expressed Basedon the dynamic measurement principle this paper proposedan optimized design method for ergonomic dimensions byGA e results indicated that the method was effective andprovided a new way for product shape innovation andhuman-machine dimensional automatic design e mainconclusions of this study were as follows

(1) Based on the information entropy theory the weightvalues of the three driving assistance functions werecalculated e results showed that the informationgain of the starting mode of the self-adaptive cruisewas highest while the information gain of the po-sition of the function key to adjust the distancebetween cars was lowest which provided a theo-retical basis for the dimensional fitness evaluationmechanism of the HMI

(2) e evolutionary process of HMI dimensionsdriven by user expectation was studied and theevolution conditions were analyzed e studyconfirmed the feasibility of normalized geometricselection arithmetic uniform crossover and non-uniform mutation in iterative researches for hu-man-machine dimensions rough parameterdebugging this evolutionary method was proved tobe effective in obtaining appropriate sizes of HMIcomponents e model could make full use of theadvantages of GA to improve the user researchperformance such as few restrictions on the opti-mization high robustness and an easily achievedglobal search

10 Complexity

(3) After finding the dimensions with the highest fitnessand applying them to the design of the HMI within-subjects experiments showed that the completiontime of five take-over tasks was significantly shorterexcept for that of Task A2 and C2is indicated thatthe information processing of drivers under the HMIwith evolved dimensions was faster which was morebeneficial to the vehicle control and take-overquality Intelligent vehicles achieved intelligent in-formation exchange with people vehicles and roadsthrough in-vehicle sensor system and informationterminals erefore the evolved dimensions couldmake the vehicle apt to respond to the environmentand have more time to fully deal with the complextraffic environment and would help to improve thetraffic efficiency therefrom

Data Availability

e data used to support the findings of this study areavailable from the corresponding author upon request

Consent

e study respected and guaranteed the subjectsrsquo right toparticipate in the study and allowed them to withdraw from thestudy unconditionally at any stage e study strictly followedthe informed consent procedure e subjects were asked tosign a consent form to inform them of the purpose processduration of the experiment matters needing attention and themethods and purpose of collecting the data so as to reducetheir anxiety and confusion In the experiment the personalsafety and health of the subjects took priority According to theusage guidelines of the driving simulator we tried to avoid anypossible injury in driving tasks e privacy of the subjects wasstrictly protected e subjects were informed of the storageuse and confidentiality measures of their personal informationand image data in advance truthfully And we would notdisclose the information to a third party or spread them on theInternet without authorization

Conflicts of Interest

e authors declare that they have no conflicts of interest

Acknowledgments

is researchwas funded by BeijingUrbanGovernance ResearchProject (grant no 20XN244) Scientific Research Program ofBeijing Education Commission (grant no KM202010009003)Yuyou Talent Support Program of North China University ofTechnology (grant no 107051360018XN012018) Tianjin YouthTalents Reserve Support Program and Chinese ErgonomicsSociety amp Kingfar Joint Research Fund for Outstanding YoungScholars (grant number CES-Kingfar-2019-001)

References

[1] S Jin J Yang E Wang and J Liu ldquoe influence of high-speed rail on ice-snow tourism in north eastern ChinardquoTourism Management vol 78 Article ID 104070 2020

[2] J Yang R Yang J Sun T Huang and Q Ge ldquoe spatialdifferentiation of the suitability of ice-snow tourist destina-tions based on a comprehensive evaluation model in ChinardquoSustainability vol 9 no 5 Article ID 774 2017

[3] W-Y Chung T-W Chong and B-G Lee ldquoMethods to detectand reduce driver stress a reviewrdquo International Journal ofAutomotive Technology vol 20 no 5 pp 1051ndash1063 2019

[4] H Farah and H N Koutsopoulos ldquoDo cooperative systemsmake driversrsquo car-following behavior saferrdquo TransportationResearch Part C Emerging Technologies vol 41 pp 61ndash722014

[5] SAE Taxonomy and Definitions for Terms Related to DrivingAutomation Systems for On-Road Motor Vehicles SAEJ3016TM Warrendale PA USA 2018

[6] J Hoegg J W Alba and D W Dahl ldquoe good the bad andthe ugly influence of aesthetics on product feature judg-mentsrdquo Journal of Consumer Psychology vol 20 no 4pp 419ndash430 2010

[7] M Kyriakidis R Happee and J C F De Winter ldquoPublicopinion on automated driving results of an internationalquestionnaire among 5000 respondentsrdquo TransportationResearch Part F Traffic Psychology and Behaviour vol 32pp 127ndash140 2015

[8] F Meng and C Spence ldquoTactile warning signals for in-vehiclesystemsrdquo Accident Analysis amp Prevention vol 75 pp 333ndash346 2015

[9] J Wan and C Wu ldquoe effects of lead time of take-overrequest and nondriving tasks on taking-over control of au-tomated vehiclesrdquo IEEE Transactions on Human MachineSystems vol 48 no 6 pp 582ndash591 2018

[10] S Petermeijer P Hornberger I Ganotis J D Winter andK Bengler ldquoe design of a vibrotactile seat for conveyingtake-over requests in automated drivingrdquo in Proceedings of the8th International Conference on Applied Human Factors andErgonomics Los Angeles CA USA July 2017

[11] T Ito A Takata and K Oosawa ldquoTime required for take-overfrom automated to manual drivingrdquo in Proceedings of the SAE2016 World Congress and Exhibition Detroit MI USA April2016

[12] K Zeeb A Buchner and M Schrauf ldquoWhat determines thetake-over time An integrated model approach of driver take-over after automated drivingrdquo Accident Analysis amp Preven-tion vol 78 pp 212ndash221 2015

[13] B Mok M Johns K J Lee et al ldquoEmergency automation offunstructured transition timing for distracted drivers of au-tomated vehiclesrdquo in Proceedings of the IEEE InternationalConference on Intelligent Transportation Systems Las PalmasSpain September 2015

[14] S Samuel A Borowsky S Zilberstein and D L FisherldquoMinimum time to situation awareness in scenarios involvingtransfer of control from an automated driving suiterdquoTransportation Research Record Journal of the TransportationResearch Board vol 2602 no 2602 pp 115ndash120 2016

[15] W Vlakveld N van Nes J de Bruin L Vissers andM van der Kroft ldquoSituation awareness increases when drivershave more time to take over the wheel in a level 3 automated

Complexity 11

car a simulator studyrdquo Transportation Research Part F TrafficPsychology and Behaviour vol 58 pp 917ndash929 2018

[16] N Merat A H Jamson F C H Lai M Daly andO M J Carsten ldquoTransition to manual driver behaviourwhen resuming control from a highly automated vehiclerdquoTransportation Research Part F Traffic Psychology and Be-haviour vol 27 no 26 pp 274ndash282 2014

[17] C Z Wu H R Wu and N C Lyu ldquoReview of control switchand safety of human-computer driving intelligent vehiclerdquoJournal of Traffic and Transportation Engineering vol 18no 6 pp 131ndash141 2018

[18] C Gold D Dambock L Lorenz and K Bengler ldquoldquoTakeoverrdquo how long does it take to get the driver back into thelooprdquo Proceedings of the Human Factors and ErgonomicsSociety Annual Meeting vol 57 no 1 pp 1938ndash1942 2013

[19] H Yang Y Zhao and Y Wang ldquoIdentifying modeling formsof instrument panel system in intelligent shared cars a studyfor perceptual preference and in-vehicle behaviorsrdquo Envi-ronmental Science and Pollution Research vol 27 no 1pp 1009ndash1023 2020

[20] A Berkovich A Lu B Levine and A V Reddy ldquoObservedcustomer seating and standing behavior and seat preferenceson board subway cars in New York cityrdquo TransportationResearch Record Journal of the Transportation ResearchBoard vol 2353 no 1 pp 33ndash46 2013

[21] R Li Y V Chen C Sha and Z Lu ldquoEffects of interface layouton the usability of in-vehicle information systems and drivingsafetyrdquo Displays vol 49 pp 124ndash132 2017

[22] H Kim S Kwon J Heo H Lee and M K Chung ldquoe effectof touch-key size on the usability of in-vehicle informationsystems and driving safety during simulated drivingrdquo AppliedErgonomics vol 45 no 3 pp 379ndash388 2014

[23] H Kim and H Song ldquoEvaluation of the safety and usability oftouch gestures in operating in-vehicle information systemswith visual occlusionrdquo Applied Ergonomics vol 45 no 3pp 789ndash798 2014

[24] M Koerber C Gold D Lechner and K Bengler ldquoe in-fluence of age on the take-over of vehicle control in highlyautomated drivingrdquo Transportation Research Part F TrafficPsychology amp Behaviour vol 39 pp 19ndash32 2016

[25] H Clark A C Mclaughlin B Williams and J Feng ldquoPer-formance in takeover and characteristics of non-driving re-lated tasks during highly automated driving in younger andolder driversrdquo Proceedings of the Human Factors amp Ergo-nomics Society Annual Meeting vol 61 no 1 pp 37ndash41 2017

[26] I Lijarcio S A Useche J Llamazares and L MontoroldquoPerceived benefits and constraints in vehicle automationdata to assess the relationship between driverrsquos features andtheir attitudes towards autonomous vehiclesrdquo Data in Briefvol 27 Article ID 104662 2019

[27] K-Z Chen and X-A Feng ldquoVirtual genes of manufacturingproducts and their reforms for product innovative designrdquoProceedings of the Institution of Mechanical Engineers Part CJournal of Mechanical Engineering Science vol 218 no 5pp 557ndash574 2004

[28] K-Z Chen X-A Feng and X-C Chen ldquoReverse deductionof virtual chromosomes of manufactured products for theirgene-engineering-based innovative designrdquo Computer-AidedDesign vol 37 no 11 pp 1191ndash1203 2005

[29] H-C Tsai and J-R Chou ldquoAutomatic design support andimage evaluation of two-coloured products using colour as-sociation and colour harmony scales and genetic algorithmrdquoComputer-Aided Design vol 39 no 9 pp 818ndash828 2007

[30] S-W Hsiao C-F Hsu and K-W Tang ldquoA consultation andsimulation system for product color planning based on in-teractive genetic algorithmsrdquo Color Research amp Applicationvol 38 no 5 pp 375ndash390 2013

[31] Y Sun W-l Wang Y-w Zhao and X-j Liu ldquoUser imageoriented interactive genetic algorithm evaluation mode inproduct developmentrdquo Computer Integrated ManufacturingSystems vol 18 no 2 pp 276ndash281 2012

[32] R Kamalian E Yeh Z Ying A M Agogino and H TakagildquoReducing human fatigue in interactive evolutionary com-putation through fuzzy systems and machine learning sys-temsrdquo in Proceedings of the 2006 IEEE InternationalConference on Fuzzy Systems Vancouver Canada July 2006

[33] Z Gu M Xi Tang and J H Frazer ldquoCapturing aestheticintention during interactive evolutionrdquo Computer-AidedDesign vol 38 no 3 pp 224ndash237 2006

[34] G Harih M Borovinsek and Z Ren Optimisation ofProductrsquos Hand-Handle Interface Material Parameters forImproved Ergonomics Springer International PublishingCham Switzerland 2015

[35] P Cheng D Chen and J Wang ldquoClustering of the bodyshape of the adult male by using principal component analysisand genetic algorithm-BP neural networkrdquo Soft Computingvol 24 no 17 Article ID 13219 2020

[36] N M B Serrano P J G Nieto A S Sanchez F S Lasherasand P R Fernandez AHybrid Algorithm for the Assessment ofthe Influence of Risk Factors in the Development of Upper LimbMusculoskeletal Disorders Springer Cham Switzerland 2018

[37] S S Sankar O-M Holman A F Gazabon and C J AcevedoldquoApplication of genetic algorithm to job scheduling underergonomic constraints in manufacturing industryrdquo Journal ofAmbient Intelligence and Humanized Computing vol 10no 5 pp 2063ndash2090 2019

[38] A-Z Atiya L Lee and K Xing ldquoDeveloping a multi-ob-jective genetic optimisation approach for an operationaldesign of a manual mixed-model assembly line with walkingworkersrdquo Journal of Intelligent Manufacturing vol 27 no 5pp 1ndash17 2014

[39] P Jedlicka and T Ryba ldquoGenetic algorithm application inimage segmentationrdquo Pattern Recognition and Image Anal-ysis vol 26 no 3 pp 497ndash501 2016

[40] L V Campenhout J Frens K Overbeeke A Standaert andH Peremans ldquoPhysical interaction in a dematerializedworldrdquo International Journal of Design vol 7 no 1 pp 1ndash182013

[41] K Nandhini and S R Balasundaram ldquoImproving readabilitythrough individualized summary extraction using interactivegenetic algorithmrdquo Applied Artificial Intelligence vol 30no 7 pp 635ndash661 2016

[42] R M Gray Entropy and Information 6eory Springer NewYork NY USA 2nd edition 2011

[43] B Anderhofstadt and S Spinler ldquoFactors affecting the pur-chasing decision and operation of alternative fuel-poweredheavy-duty trucks in Germanymdasha Delphi studyrdquo Trans-portation Research Part D Transport and Environmentvol 73 pp 87ndash107 2019

[44] X-m Wu and Y Feng ldquoReactive power optimization ofdistribution network based on improved genetic algorithmrdquoJournal of Xirsquoan Shiyou University (Natural Science Edition)vol 30 no 3 pp 95ndash99 2015

[45] Z Ling andM-j Li ldquoState space evolutionary algorithm basedon non-uniform mutation operatorrdquo Computer Technologyand Development vol 28 no 9 pp 68ndash71 2018

12 Complexity

[46] W Hu and J Zhao ldquoAutomobile styling gene evolutiondriven by usersrsquo expectation imagerdquo Journal of MechanicalEngineering vol 47 no 16 pp 176ndash181 2011

[47] J J Scott and R Gray ldquoA comparison of tactile visual andauditory warnings for rear-end collision prevention in sim-ulated drivingrdquo Human Factors 6e Journal of the HumanFactors and Ergonomics Society vol 50 no 2 pp 264ndash2752008

[48] B Reimer B Mehler and J F Coughlin ldquoReductions in self-reported stress and anticipatory heart rate with the use of asemi-automated parallel parking systemrdquo Applied Ergonom-ics vol 52 pp 120ndash127 2016

[49] B DrsquoIppolito ldquoe importance of design for firms com-petitiveness a review of the literaturerdquo Technovation vol 34no 11 pp 716ndash730 2014

[50] S Asensio-Cuesta J A Diego-Mas L Canos-Daros andC Andres-Romano ldquoA genetic algorithm for the design of jobrotation schedules considering ergonomic and competencecriteriardquo International Journal of Advanced ManufacturingTechnology vol 60 no 9ndash12 pp 1161ndash1174 2012

[51] S-C Lee H-E Tseng C-C Chang and Y-M HuangldquoApplying interactive genetic algorithms to disassembly se-quence planningrdquo International Journal of Precision Engi-neering and Manufacturing vol 21 no 4 pp 663ndash679 2020

[52] M G Shishaev V V Dikovitsky and L V Lapochkina ldquoeexperience of building cognitive user interfaces of multido-main information systems based on the mental model ofusersrdquo in Proceedings of the 6th Computer Science On-LineConference (CSOC 2017) Prague Czech Republic April 2017

[53] S Lee H Alzoubi and S Kim ldquoe effect of interior designelements and lighting layouts on prospective occupantsrsquoperceptions of amenity and efficiency in living roomsrdquo Sus-tainability vol 9 no 7 Article ID 1119 2017

[54] G Caruso ldquoMixed reality system for ergonomic assessment ofdriverrsquos seatrdquo International Journal of Virtual Reality vol 10no 2 pp 69ndash79 2011

[55] M L Reyes and J D Lee ldquoEffects of cognitive load presenceand duration on driver eye movements and event detectionperformancerdquo Transportation Research Part F Traffic Psy-chology and Behaviour vol 11 no 6 pp 391ndash402 2008

[56] D Cisler P M Greenwood D M Roberts R McKendrickand C L Baldwin ldquoComparing the relative strengths of EEGand low-cost physiological devices in modeling attentionallocation in semiautonomous vehiclesrdquo Frontiers in HumanNeuroscience vol 13 Article ID 109 2019

[57] I-H Kim J-W Kim S Haufe and S-W Lee ldquoDetection ofbraking intention in diverse situations during simulateddriving based on EEG feature combinationrdquo Journal of NeuralEngineering vol 12 no 1 Article ID 016001 2014

Complexity 13

(3) After finding the dimensions with the highest fitnessand applying them to the design of the HMI within-subjects experiments showed that the completiontime of five take-over tasks was significantly shorterexcept for that of Task A2 and C2is indicated thatthe information processing of drivers under the HMIwith evolved dimensions was faster which was morebeneficial to the vehicle control and take-overquality Intelligent vehicles achieved intelligent in-formation exchange with people vehicles and roadsthrough in-vehicle sensor system and informationterminals erefore the evolved dimensions couldmake the vehicle apt to respond to the environmentand have more time to fully deal with the complextraffic environment and would help to improve thetraffic efficiency therefrom

Data Availability

e data used to support the findings of this study areavailable from the corresponding author upon request

Consent

e study respected and guaranteed the subjectsrsquo right toparticipate in the study and allowed them to withdraw from thestudy unconditionally at any stage e study strictly followedthe informed consent procedure e subjects were asked tosign a consent form to inform them of the purpose processduration of the experiment matters needing attention and themethods and purpose of collecting the data so as to reducetheir anxiety and confusion In the experiment the personalsafety and health of the subjects took priority According to theusage guidelines of the driving simulator we tried to avoid anypossible injury in driving tasks e privacy of the subjects wasstrictly protected e subjects were informed of the storageuse and confidentiality measures of their personal informationand image data in advance truthfully And we would notdisclose the information to a third party or spread them on theInternet without authorization

Conflicts of Interest

e authors declare that they have no conflicts of interest

Acknowledgments

is researchwas funded by BeijingUrbanGovernance ResearchProject (grant no 20XN244) Scientific Research Program ofBeijing Education Commission (grant no KM202010009003)Yuyou Talent Support Program of North China University ofTechnology (grant no 107051360018XN012018) Tianjin YouthTalents Reserve Support Program and Chinese ErgonomicsSociety amp Kingfar Joint Research Fund for Outstanding YoungScholars (grant number CES-Kingfar-2019-001)

References

[1] S Jin J Yang E Wang and J Liu ldquoe influence of high-speed rail on ice-snow tourism in north eastern ChinardquoTourism Management vol 78 Article ID 104070 2020

[2] J Yang R Yang J Sun T Huang and Q Ge ldquoe spatialdifferentiation of the suitability of ice-snow tourist destina-tions based on a comprehensive evaluation model in ChinardquoSustainability vol 9 no 5 Article ID 774 2017

[3] W-Y Chung T-W Chong and B-G Lee ldquoMethods to detectand reduce driver stress a reviewrdquo International Journal ofAutomotive Technology vol 20 no 5 pp 1051ndash1063 2019

[4] H Farah and H N Koutsopoulos ldquoDo cooperative systemsmake driversrsquo car-following behavior saferrdquo TransportationResearch Part C Emerging Technologies vol 41 pp 61ndash722014

[5] SAE Taxonomy and Definitions for Terms Related to DrivingAutomation Systems for On-Road Motor Vehicles SAEJ3016TM Warrendale PA USA 2018

[6] J Hoegg J W Alba and D W Dahl ldquoe good the bad andthe ugly influence of aesthetics on product feature judg-mentsrdquo Journal of Consumer Psychology vol 20 no 4pp 419ndash430 2010

[7] M Kyriakidis R Happee and J C F De Winter ldquoPublicopinion on automated driving results of an internationalquestionnaire among 5000 respondentsrdquo TransportationResearch Part F Traffic Psychology and Behaviour vol 32pp 127ndash140 2015

[8] F Meng and C Spence ldquoTactile warning signals for in-vehiclesystemsrdquo Accident Analysis amp Prevention vol 75 pp 333ndash346 2015

[9] J Wan and C Wu ldquoe effects of lead time of take-overrequest and nondriving tasks on taking-over control of au-tomated vehiclesrdquo IEEE Transactions on Human MachineSystems vol 48 no 6 pp 582ndash591 2018

[10] S Petermeijer P Hornberger I Ganotis J D Winter andK Bengler ldquoe design of a vibrotactile seat for conveyingtake-over requests in automated drivingrdquo in Proceedings of the8th International Conference on Applied Human Factors andErgonomics Los Angeles CA USA July 2017

[11] T Ito A Takata and K Oosawa ldquoTime required for take-overfrom automated to manual drivingrdquo in Proceedings of the SAE2016 World Congress and Exhibition Detroit MI USA April2016

[12] K Zeeb A Buchner and M Schrauf ldquoWhat determines thetake-over time An integrated model approach of driver take-over after automated drivingrdquo Accident Analysis amp Preven-tion vol 78 pp 212ndash221 2015

[13] B Mok M Johns K J Lee et al ldquoEmergency automation offunstructured transition timing for distracted drivers of au-tomated vehiclesrdquo in Proceedings of the IEEE InternationalConference on Intelligent Transportation Systems Las PalmasSpain September 2015

[14] S Samuel A Borowsky S Zilberstein and D L FisherldquoMinimum time to situation awareness in scenarios involvingtransfer of control from an automated driving suiterdquoTransportation Research Record Journal of the TransportationResearch Board vol 2602 no 2602 pp 115ndash120 2016

[15] W Vlakveld N van Nes J de Bruin L Vissers andM van der Kroft ldquoSituation awareness increases when drivershave more time to take over the wheel in a level 3 automated

Complexity 11

car a simulator studyrdquo Transportation Research Part F TrafficPsychology and Behaviour vol 58 pp 917ndash929 2018

[16] N Merat A H Jamson F C H Lai M Daly andO M J Carsten ldquoTransition to manual driver behaviourwhen resuming control from a highly automated vehiclerdquoTransportation Research Part F Traffic Psychology and Be-haviour vol 27 no 26 pp 274ndash282 2014

[17] C Z Wu H R Wu and N C Lyu ldquoReview of control switchand safety of human-computer driving intelligent vehiclerdquoJournal of Traffic and Transportation Engineering vol 18no 6 pp 131ndash141 2018

[18] C Gold D Dambock L Lorenz and K Bengler ldquoldquoTakeoverrdquo how long does it take to get the driver back into thelooprdquo Proceedings of the Human Factors and ErgonomicsSociety Annual Meeting vol 57 no 1 pp 1938ndash1942 2013

[19] H Yang Y Zhao and Y Wang ldquoIdentifying modeling formsof instrument panel system in intelligent shared cars a studyfor perceptual preference and in-vehicle behaviorsrdquo Envi-ronmental Science and Pollution Research vol 27 no 1pp 1009ndash1023 2020

[20] A Berkovich A Lu B Levine and A V Reddy ldquoObservedcustomer seating and standing behavior and seat preferenceson board subway cars in New York cityrdquo TransportationResearch Record Journal of the Transportation ResearchBoard vol 2353 no 1 pp 33ndash46 2013

[21] R Li Y V Chen C Sha and Z Lu ldquoEffects of interface layouton the usability of in-vehicle information systems and drivingsafetyrdquo Displays vol 49 pp 124ndash132 2017

[22] H Kim S Kwon J Heo H Lee and M K Chung ldquoe effectof touch-key size on the usability of in-vehicle informationsystems and driving safety during simulated drivingrdquo AppliedErgonomics vol 45 no 3 pp 379ndash388 2014

[23] H Kim and H Song ldquoEvaluation of the safety and usability oftouch gestures in operating in-vehicle information systemswith visual occlusionrdquo Applied Ergonomics vol 45 no 3pp 789ndash798 2014

[24] M Koerber C Gold D Lechner and K Bengler ldquoe in-fluence of age on the take-over of vehicle control in highlyautomated drivingrdquo Transportation Research Part F TrafficPsychology amp Behaviour vol 39 pp 19ndash32 2016

[25] H Clark A C Mclaughlin B Williams and J Feng ldquoPer-formance in takeover and characteristics of non-driving re-lated tasks during highly automated driving in younger andolder driversrdquo Proceedings of the Human Factors amp Ergo-nomics Society Annual Meeting vol 61 no 1 pp 37ndash41 2017

[26] I Lijarcio S A Useche J Llamazares and L MontoroldquoPerceived benefits and constraints in vehicle automationdata to assess the relationship between driverrsquos features andtheir attitudes towards autonomous vehiclesrdquo Data in Briefvol 27 Article ID 104662 2019

[27] K-Z Chen and X-A Feng ldquoVirtual genes of manufacturingproducts and their reforms for product innovative designrdquoProceedings of the Institution of Mechanical Engineers Part CJournal of Mechanical Engineering Science vol 218 no 5pp 557ndash574 2004

[28] K-Z Chen X-A Feng and X-C Chen ldquoReverse deductionof virtual chromosomes of manufactured products for theirgene-engineering-based innovative designrdquo Computer-AidedDesign vol 37 no 11 pp 1191ndash1203 2005

[29] H-C Tsai and J-R Chou ldquoAutomatic design support andimage evaluation of two-coloured products using colour as-sociation and colour harmony scales and genetic algorithmrdquoComputer-Aided Design vol 39 no 9 pp 818ndash828 2007

[30] S-W Hsiao C-F Hsu and K-W Tang ldquoA consultation andsimulation system for product color planning based on in-teractive genetic algorithmsrdquo Color Research amp Applicationvol 38 no 5 pp 375ndash390 2013

[31] Y Sun W-l Wang Y-w Zhao and X-j Liu ldquoUser imageoriented interactive genetic algorithm evaluation mode inproduct developmentrdquo Computer Integrated ManufacturingSystems vol 18 no 2 pp 276ndash281 2012

[32] R Kamalian E Yeh Z Ying A M Agogino and H TakagildquoReducing human fatigue in interactive evolutionary com-putation through fuzzy systems and machine learning sys-temsrdquo in Proceedings of the 2006 IEEE InternationalConference on Fuzzy Systems Vancouver Canada July 2006

[33] Z Gu M Xi Tang and J H Frazer ldquoCapturing aestheticintention during interactive evolutionrdquo Computer-AidedDesign vol 38 no 3 pp 224ndash237 2006

[34] G Harih M Borovinsek and Z Ren Optimisation ofProductrsquos Hand-Handle Interface Material Parameters forImproved Ergonomics Springer International PublishingCham Switzerland 2015

[35] P Cheng D Chen and J Wang ldquoClustering of the bodyshape of the adult male by using principal component analysisand genetic algorithm-BP neural networkrdquo Soft Computingvol 24 no 17 Article ID 13219 2020

[36] N M B Serrano P J G Nieto A S Sanchez F S Lasherasand P R Fernandez AHybrid Algorithm for the Assessment ofthe Influence of Risk Factors in the Development of Upper LimbMusculoskeletal Disorders Springer Cham Switzerland 2018

[37] S S Sankar O-M Holman A F Gazabon and C J AcevedoldquoApplication of genetic algorithm to job scheduling underergonomic constraints in manufacturing industryrdquo Journal ofAmbient Intelligence and Humanized Computing vol 10no 5 pp 2063ndash2090 2019

[38] A-Z Atiya L Lee and K Xing ldquoDeveloping a multi-ob-jective genetic optimisation approach for an operationaldesign of a manual mixed-model assembly line with walkingworkersrdquo Journal of Intelligent Manufacturing vol 27 no 5pp 1ndash17 2014

[39] P Jedlicka and T Ryba ldquoGenetic algorithm application inimage segmentationrdquo Pattern Recognition and Image Anal-ysis vol 26 no 3 pp 497ndash501 2016

[40] L V Campenhout J Frens K Overbeeke A Standaert andH Peremans ldquoPhysical interaction in a dematerializedworldrdquo International Journal of Design vol 7 no 1 pp 1ndash182013

[41] K Nandhini and S R Balasundaram ldquoImproving readabilitythrough individualized summary extraction using interactivegenetic algorithmrdquo Applied Artificial Intelligence vol 30no 7 pp 635ndash661 2016

[42] R M Gray Entropy and Information 6eory Springer NewYork NY USA 2nd edition 2011

[43] B Anderhofstadt and S Spinler ldquoFactors affecting the pur-chasing decision and operation of alternative fuel-poweredheavy-duty trucks in Germanymdasha Delphi studyrdquo Trans-portation Research Part D Transport and Environmentvol 73 pp 87ndash107 2019

[44] X-m Wu and Y Feng ldquoReactive power optimization ofdistribution network based on improved genetic algorithmrdquoJournal of Xirsquoan Shiyou University (Natural Science Edition)vol 30 no 3 pp 95ndash99 2015

[45] Z Ling andM-j Li ldquoState space evolutionary algorithm basedon non-uniform mutation operatorrdquo Computer Technologyand Development vol 28 no 9 pp 68ndash71 2018

12 Complexity

[46] W Hu and J Zhao ldquoAutomobile styling gene evolutiondriven by usersrsquo expectation imagerdquo Journal of MechanicalEngineering vol 47 no 16 pp 176ndash181 2011

[47] J J Scott and R Gray ldquoA comparison of tactile visual andauditory warnings for rear-end collision prevention in sim-ulated drivingrdquo Human Factors 6e Journal of the HumanFactors and Ergonomics Society vol 50 no 2 pp 264ndash2752008

[48] B Reimer B Mehler and J F Coughlin ldquoReductions in self-reported stress and anticipatory heart rate with the use of asemi-automated parallel parking systemrdquo Applied Ergonom-ics vol 52 pp 120ndash127 2016

[49] B DrsquoIppolito ldquoe importance of design for firms com-petitiveness a review of the literaturerdquo Technovation vol 34no 11 pp 716ndash730 2014

[50] S Asensio-Cuesta J A Diego-Mas L Canos-Daros andC Andres-Romano ldquoA genetic algorithm for the design of jobrotation schedules considering ergonomic and competencecriteriardquo International Journal of Advanced ManufacturingTechnology vol 60 no 9ndash12 pp 1161ndash1174 2012

[51] S-C Lee H-E Tseng C-C Chang and Y-M HuangldquoApplying interactive genetic algorithms to disassembly se-quence planningrdquo International Journal of Precision Engi-neering and Manufacturing vol 21 no 4 pp 663ndash679 2020

[52] M G Shishaev V V Dikovitsky and L V Lapochkina ldquoeexperience of building cognitive user interfaces of multido-main information systems based on the mental model ofusersrdquo in Proceedings of the 6th Computer Science On-LineConference (CSOC 2017) Prague Czech Republic April 2017

[53] S Lee H Alzoubi and S Kim ldquoe effect of interior designelements and lighting layouts on prospective occupantsrsquoperceptions of amenity and efficiency in living roomsrdquo Sus-tainability vol 9 no 7 Article ID 1119 2017

[54] G Caruso ldquoMixed reality system for ergonomic assessment ofdriverrsquos seatrdquo International Journal of Virtual Reality vol 10no 2 pp 69ndash79 2011

[55] M L Reyes and J D Lee ldquoEffects of cognitive load presenceand duration on driver eye movements and event detectionperformancerdquo Transportation Research Part F Traffic Psy-chology and Behaviour vol 11 no 6 pp 391ndash402 2008

[56] D Cisler P M Greenwood D M Roberts R McKendrickand C L Baldwin ldquoComparing the relative strengths of EEGand low-cost physiological devices in modeling attentionallocation in semiautonomous vehiclesrdquo Frontiers in HumanNeuroscience vol 13 Article ID 109 2019

[57] I-H Kim J-W Kim S Haufe and S-W Lee ldquoDetection ofbraking intention in diverse situations during simulateddriving based on EEG feature combinationrdquo Journal of NeuralEngineering vol 12 no 1 Article ID 016001 2014

Complexity 13

car a simulator studyrdquo Transportation Research Part F TrafficPsychology and Behaviour vol 58 pp 917ndash929 2018

[16] N Merat A H Jamson F C H Lai M Daly andO M J Carsten ldquoTransition to manual driver behaviourwhen resuming control from a highly automated vehiclerdquoTransportation Research Part F Traffic Psychology and Be-haviour vol 27 no 26 pp 274ndash282 2014

[17] C Z Wu H R Wu and N C Lyu ldquoReview of control switchand safety of human-computer driving intelligent vehiclerdquoJournal of Traffic and Transportation Engineering vol 18no 6 pp 131ndash141 2018

[18] C Gold D Dambock L Lorenz and K Bengler ldquoldquoTakeoverrdquo how long does it take to get the driver back into thelooprdquo Proceedings of the Human Factors and ErgonomicsSociety Annual Meeting vol 57 no 1 pp 1938ndash1942 2013

[19] H Yang Y Zhao and Y Wang ldquoIdentifying modeling formsof instrument panel system in intelligent shared cars a studyfor perceptual preference and in-vehicle behaviorsrdquo Envi-ronmental Science and Pollution Research vol 27 no 1pp 1009ndash1023 2020

[20] A Berkovich A Lu B Levine and A V Reddy ldquoObservedcustomer seating and standing behavior and seat preferenceson board subway cars in New York cityrdquo TransportationResearch Record Journal of the Transportation ResearchBoard vol 2353 no 1 pp 33ndash46 2013

[21] R Li Y V Chen C Sha and Z Lu ldquoEffects of interface layouton the usability of in-vehicle information systems and drivingsafetyrdquo Displays vol 49 pp 124ndash132 2017

[22] H Kim S Kwon J Heo H Lee and M K Chung ldquoe effectof touch-key size on the usability of in-vehicle informationsystems and driving safety during simulated drivingrdquo AppliedErgonomics vol 45 no 3 pp 379ndash388 2014

[23] H Kim and H Song ldquoEvaluation of the safety and usability oftouch gestures in operating in-vehicle information systemswith visual occlusionrdquo Applied Ergonomics vol 45 no 3pp 789ndash798 2014

[24] M Koerber C Gold D Lechner and K Bengler ldquoe in-fluence of age on the take-over of vehicle control in highlyautomated drivingrdquo Transportation Research Part F TrafficPsychology amp Behaviour vol 39 pp 19ndash32 2016

[25] H Clark A C Mclaughlin B Williams and J Feng ldquoPer-formance in takeover and characteristics of non-driving re-lated tasks during highly automated driving in younger andolder driversrdquo Proceedings of the Human Factors amp Ergo-nomics Society Annual Meeting vol 61 no 1 pp 37ndash41 2017

[26] I Lijarcio S A Useche J Llamazares and L MontoroldquoPerceived benefits and constraints in vehicle automationdata to assess the relationship between driverrsquos features andtheir attitudes towards autonomous vehiclesrdquo Data in Briefvol 27 Article ID 104662 2019

[27] K-Z Chen and X-A Feng ldquoVirtual genes of manufacturingproducts and their reforms for product innovative designrdquoProceedings of the Institution of Mechanical Engineers Part CJournal of Mechanical Engineering Science vol 218 no 5pp 557ndash574 2004

[28] K-Z Chen X-A Feng and X-C Chen ldquoReverse deductionof virtual chromosomes of manufactured products for theirgene-engineering-based innovative designrdquo Computer-AidedDesign vol 37 no 11 pp 1191ndash1203 2005

[29] H-C Tsai and J-R Chou ldquoAutomatic design support andimage evaluation of two-coloured products using colour as-sociation and colour harmony scales and genetic algorithmrdquoComputer-Aided Design vol 39 no 9 pp 818ndash828 2007

[30] S-W Hsiao C-F Hsu and K-W Tang ldquoA consultation andsimulation system for product color planning based on in-teractive genetic algorithmsrdquo Color Research amp Applicationvol 38 no 5 pp 375ndash390 2013

[31] Y Sun W-l Wang Y-w Zhao and X-j Liu ldquoUser imageoriented interactive genetic algorithm evaluation mode inproduct developmentrdquo Computer Integrated ManufacturingSystems vol 18 no 2 pp 276ndash281 2012

[32] R Kamalian E Yeh Z Ying A M Agogino and H TakagildquoReducing human fatigue in interactive evolutionary com-putation through fuzzy systems and machine learning sys-temsrdquo in Proceedings of the 2006 IEEE InternationalConference on Fuzzy Systems Vancouver Canada July 2006

[33] Z Gu M Xi Tang and J H Frazer ldquoCapturing aestheticintention during interactive evolutionrdquo Computer-AidedDesign vol 38 no 3 pp 224ndash237 2006

[34] G Harih M Borovinsek and Z Ren Optimisation ofProductrsquos Hand-Handle Interface Material Parameters forImproved Ergonomics Springer International PublishingCham Switzerland 2015

[35] P Cheng D Chen and J Wang ldquoClustering of the bodyshape of the adult male by using principal component analysisand genetic algorithm-BP neural networkrdquo Soft Computingvol 24 no 17 Article ID 13219 2020

[36] N M B Serrano P J G Nieto A S Sanchez F S Lasherasand P R Fernandez AHybrid Algorithm for the Assessment ofthe Influence of Risk Factors in the Development of Upper LimbMusculoskeletal Disorders Springer Cham Switzerland 2018

[37] S S Sankar O-M Holman A F Gazabon and C J AcevedoldquoApplication of genetic algorithm to job scheduling underergonomic constraints in manufacturing industryrdquo Journal ofAmbient Intelligence and Humanized Computing vol 10no 5 pp 2063ndash2090 2019

[38] A-Z Atiya L Lee and K Xing ldquoDeveloping a multi-ob-jective genetic optimisation approach for an operationaldesign of a manual mixed-model assembly line with walkingworkersrdquo Journal of Intelligent Manufacturing vol 27 no 5pp 1ndash17 2014

[39] P Jedlicka and T Ryba ldquoGenetic algorithm application inimage segmentationrdquo Pattern Recognition and Image Anal-ysis vol 26 no 3 pp 497ndash501 2016

[40] L V Campenhout J Frens K Overbeeke A Standaert andH Peremans ldquoPhysical interaction in a dematerializedworldrdquo International Journal of Design vol 7 no 1 pp 1ndash182013

[41] K Nandhini and S R Balasundaram ldquoImproving readabilitythrough individualized summary extraction using interactivegenetic algorithmrdquo Applied Artificial Intelligence vol 30no 7 pp 635ndash661 2016

[42] R M Gray Entropy and Information 6eory Springer NewYork NY USA 2nd edition 2011

[43] B Anderhofstadt and S Spinler ldquoFactors affecting the pur-chasing decision and operation of alternative fuel-poweredheavy-duty trucks in Germanymdasha Delphi studyrdquo Trans-portation Research Part D Transport and Environmentvol 73 pp 87ndash107 2019

[44] X-m Wu and Y Feng ldquoReactive power optimization ofdistribution network based on improved genetic algorithmrdquoJournal of Xirsquoan Shiyou University (Natural Science Edition)vol 30 no 3 pp 95ndash99 2015

[45] Z Ling andM-j Li ldquoState space evolutionary algorithm basedon non-uniform mutation operatorrdquo Computer Technologyand Development vol 28 no 9 pp 68ndash71 2018

12 Complexity

[46] W Hu and J Zhao ldquoAutomobile styling gene evolutiondriven by usersrsquo expectation imagerdquo Journal of MechanicalEngineering vol 47 no 16 pp 176ndash181 2011

[47] J J Scott and R Gray ldquoA comparison of tactile visual andauditory warnings for rear-end collision prevention in sim-ulated drivingrdquo Human Factors 6e Journal of the HumanFactors and Ergonomics Society vol 50 no 2 pp 264ndash2752008

[48] B Reimer B Mehler and J F Coughlin ldquoReductions in self-reported stress and anticipatory heart rate with the use of asemi-automated parallel parking systemrdquo Applied Ergonom-ics vol 52 pp 120ndash127 2016

[49] B DrsquoIppolito ldquoe importance of design for firms com-petitiveness a review of the literaturerdquo Technovation vol 34no 11 pp 716ndash730 2014

[50] S Asensio-Cuesta J A Diego-Mas L Canos-Daros andC Andres-Romano ldquoA genetic algorithm for the design of jobrotation schedules considering ergonomic and competencecriteriardquo International Journal of Advanced ManufacturingTechnology vol 60 no 9ndash12 pp 1161ndash1174 2012

[51] S-C Lee H-E Tseng C-C Chang and Y-M HuangldquoApplying interactive genetic algorithms to disassembly se-quence planningrdquo International Journal of Precision Engi-neering and Manufacturing vol 21 no 4 pp 663ndash679 2020

[52] M G Shishaev V V Dikovitsky and L V Lapochkina ldquoeexperience of building cognitive user interfaces of multido-main information systems based on the mental model ofusersrdquo in Proceedings of the 6th Computer Science On-LineConference (CSOC 2017) Prague Czech Republic April 2017

[53] S Lee H Alzoubi and S Kim ldquoe effect of interior designelements and lighting layouts on prospective occupantsrsquoperceptions of amenity and efficiency in living roomsrdquo Sus-tainability vol 9 no 7 Article ID 1119 2017

[54] G Caruso ldquoMixed reality system for ergonomic assessment ofdriverrsquos seatrdquo International Journal of Virtual Reality vol 10no 2 pp 69ndash79 2011

[55] M L Reyes and J D Lee ldquoEffects of cognitive load presenceand duration on driver eye movements and event detectionperformancerdquo Transportation Research Part F Traffic Psy-chology and Behaviour vol 11 no 6 pp 391ndash402 2008

[56] D Cisler P M Greenwood D M Roberts R McKendrickand C L Baldwin ldquoComparing the relative strengths of EEGand low-cost physiological devices in modeling attentionallocation in semiautonomous vehiclesrdquo Frontiers in HumanNeuroscience vol 13 Article ID 109 2019

[57] I-H Kim J-W Kim S Haufe and S-W Lee ldquoDetection ofbraking intention in diverse situations during simulateddriving based on EEG feature combinationrdquo Journal of NeuralEngineering vol 12 no 1 Article ID 016001 2014

Complexity 13

[46] W Hu and J Zhao ldquoAutomobile styling gene evolutiondriven by usersrsquo expectation imagerdquo Journal of MechanicalEngineering vol 47 no 16 pp 176ndash181 2011

[47] J J Scott and R Gray ldquoA comparison of tactile visual andauditory warnings for rear-end collision prevention in sim-ulated drivingrdquo Human Factors 6e Journal of the HumanFactors and Ergonomics Society vol 50 no 2 pp 264ndash2752008

[48] B Reimer B Mehler and J F Coughlin ldquoReductions in self-reported stress and anticipatory heart rate with the use of asemi-automated parallel parking systemrdquo Applied Ergonom-ics vol 52 pp 120ndash127 2016

[49] B DrsquoIppolito ldquoe importance of design for firms com-petitiveness a review of the literaturerdquo Technovation vol 34no 11 pp 716ndash730 2014

[50] S Asensio-Cuesta J A Diego-Mas L Canos-Daros andC Andres-Romano ldquoA genetic algorithm for the design of jobrotation schedules considering ergonomic and competencecriteriardquo International Journal of Advanced ManufacturingTechnology vol 60 no 9ndash12 pp 1161ndash1174 2012

[51] S-C Lee H-E Tseng C-C Chang and Y-M HuangldquoApplying interactive genetic algorithms to disassembly se-quence planningrdquo International Journal of Precision Engi-neering and Manufacturing vol 21 no 4 pp 663ndash679 2020

[52] M G Shishaev V V Dikovitsky and L V Lapochkina ldquoeexperience of building cognitive user interfaces of multido-main information systems based on the mental model ofusersrdquo in Proceedings of the 6th Computer Science On-LineConference (CSOC 2017) Prague Czech Republic April 2017

[53] S Lee H Alzoubi and S Kim ldquoe effect of interior designelements and lighting layouts on prospective occupantsrsquoperceptions of amenity and efficiency in living roomsrdquo Sus-tainability vol 9 no 7 Article ID 1119 2017

[54] G Caruso ldquoMixed reality system for ergonomic assessment ofdriverrsquos seatrdquo International Journal of Virtual Reality vol 10no 2 pp 69ndash79 2011

[55] M L Reyes and J D Lee ldquoEffects of cognitive load presenceand duration on driver eye movements and event detectionperformancerdquo Transportation Research Part F Traffic Psy-chology and Behaviour vol 11 no 6 pp 391ndash402 2008

[56] D Cisler P M Greenwood D M Roberts R McKendrickand C L Baldwin ldquoComparing the relative strengths of EEGand low-cost physiological devices in modeling attentionallocation in semiautonomous vehiclesrdquo Frontiers in HumanNeuroscience vol 13 Article ID 109 2019

[57] I-H Kim J-W Kim S Haufe and S-W Lee ldquoDetection ofbraking intention in diverse situations during simulateddriving based on EEG feature combinationrdquo Journal of NeuralEngineering vol 12 no 1 Article ID 016001 2014

Complexity 13