effects of automated traffic enforcement on driver …€¦ · saad al jaman* department of civil...

http://www.iaeme.com/IJCIET/index.asp 389 [email protected]

International Journal of Civil Engineering and Technology (IJCIET) Volume 9, Issue 2, February 2018, pp. 389–401, Article ID: IJCIET_09_02_037 Available online at http://www.iaeme.com/ijciet/issues.asp?JType=IJCIET&VType=9&IType=2 ISSN Print: 0976-6308 and ISSN Online: 0976-6316 © IAEME Publication Scopus Indexed

EFFECTS OF AUTOMATED TRAFFIC

ENFORCEMENT ON DRIVER BEHAVIOR AT A

SIGNALIZED INTERSECTION IN SAUDI

ARABIA

Saad Al Jaman*

Department of Civil Engineering, King Abdulaziz University, Saudi Arabia

MJB Alam

Department of Civil Engineering, King Abdulaziz University, Saudi Arabia

A. Hamdi

Department of Civil Engineering, King Abdulaziz University, Saudi Arabia;

*Corresponding Author

ABSTRACT:

For automating road network Saudi Arabia recently introduced 'Saher Traffic

Monitoring Enforcement System' (STMES) using red-light and speed-limit camera.

This paper analyzes the impact of STMES on driver behavior and decision making

process at signalized intersections. It proposes models for analyzing driver's

probabilities of stopping and crossing decisions at signalized intersections during

yellow-light interval on the basis of speed, distance from intersection and vehicle type.

Models also compare STMES fitted intersections with conventional intersections

without automated systems. Binary logistic regression and Artificial Neural Network

(ANN) models are developed using video data at intersections in Makkah city.

Analyses reveal that STMES increases stopping probability by 26 percent and

distance to the stop line is the most significant factor affecting the drivers’ decisions in

the yellow interval. Dilemma zone shifts closer to the stop line and the number of early

stops increases due to STMES. Artificial neural network model performs better than

binary logistic regression model in predicting the stopping probability at signalized

intersections. Research findings suggest that ANN models can be effectively used in

traffic signal design and dilemma zone reduction to improve efficiency and safety at

roadway intersections.

Keywords: Traffic, Enforcement, Dilemma, Intersection, Modeling, Logistic Regression, Neural Network

Effects of Automated Traffic Enforcement on Driver Behavior at a Signalized Intersection in Saudi Arabia


Cite this Article: Saad Al Jaman, MJB Alam amd A. Hamdi, Effects of Automated Traffic Enforcement on Driver Behavior at a Signalized Intersection in Saudi Arabia, International Journal of Civil Engineering and Technology, 9(2), 2018, pp. 389–401. http://www.iaeme.com/IJCIET/issues.asp?JType=IJCIET&VType=9&IType=2

1. INTRODUCTION

Intersections are the most critical component of urban roadway network systems in terms of performance and safety concerns. In 2008, the National Highway Traffic Safety Administration (NHTSA) estimated that crashes at signalized intersections in the USA caused more than 7,770 fatalities, 10 percent of which were caused by red-light running, and 97 percent of drivers feel that violation of red-lights at signalized intersections is a major safety concern [1]. In 2009, statistics indicated that traffic accidents in the Kingdom of Saudi Arabia led to 6,142 deaths, and red-light violations were a major cause of such incidents [2].

Red-light cameras (RLCs) and speed cameras at intersections are the most commonly used automated traffic enforcement technologies [3]. Urban areas are at a higher risk of crashes caused by the driver running a red light; approximately 22% of crashes at signalized intersections are caused by red-light violations [4, 5]. Many researchers recommend automated enforcement systems together with RLCs to improve safety at intersections [6, 5, 7]. Most researchers have focused on either crash severity (fatality, injury, vehicle damage, etc.), violation type (red-light running, speeding, etc.), or collision type (right-angle, rear-end, etc.). Overall, RLCs increase the proportion of rear-end crashes [8, 9] and decrease the proportion of right-angle crashes and red-light running crashes [10, 11].

In April 2010, the government of Saudi Arabia launched the automated Saher Traffic Monitoring Enforcement System (STMES), which controls and manages vehicular traffic. STMES is intended to reduce accidents and raise the overall efficiency of the road networks in Saudi Arabia. The system uses RLCs to enforce traffic laws at signalized intersections, particularly focusing on red-light signal violations. Understanding the drivers’ perception of STMES and the way it affects decision-making are significant factors to assess and enhance the system’s performance. According to studies published before STMES was implemented, drivers in Riyadh tended to be more aggressive than drivers in developed countries [12]. Driving behavior in terms of headway and gap acceptance in Riyadh and Makkah is also different, i.e., is shorter, from that of other countries [13, 14]. Although models for the probability of stopping at signalized intersections are often debated in international literature, no models that we are aware of, take local data from Saudi drivers into account.

This paper presents major findings of research on drivers’ perceptions and responses toward STMES and the resulting effect on traffic parameters. These results will assist in identifying the dilemma zone and devising ways to mitigate its effect to reduce traffic accidents. The study will also assist to develop a model for traffic characteristics and driver reaction patterns in Saudi Arabia to better understand the effect of STMES on driver behavior. After the literature review, the subsequent sections of this paper explain the analytical framework, results, discussion, and conclusion.

2. LITERATURE REVIEW

The study of drivers’ behavioral responses to traffic control devices is a classic research area in traffic engineering. Driver behavior influences the parameters that control intersections and determine the length of the dilemma zone. In addition, modeling of driver behavior can improve the safety of signalized intersections [15].

Most of the research on drivers’ behavior relates to accidents because human decisions are considered to be a contributing factor in more than 90% of total road accidents [16, 17].

Saad Al Jaman, MJB Alam amd A. Hamdi


NHTSA has indicated that about 67% of fatal accidents in the USA in the years 1990–1996 were caused by aggressive driving behavior [18]. Therefore, drivers’ behavior at signalized intersections, and particularly aggressive driving, may have significant implications on signal design parameters. Implementation of an enforcement system, such as the use of RLCs, complicates these scenarios and demands rigorous investigation.

Lum and Wong [19, 20, 21] conducted studies on the effect of signal-enforcement systems in Singapore on driver decisions regarding red-light signals. They collected data using loop sensors to and employed logistic modeling to model the stopping/crossing decisions of drivers while responding to yellow signals. The results indicated that RLCs significantly affect driver decisions at signalized intersections.

Papaioannou [22] investigated driving behavior at yellow signals as a contributing factor in safety at a signalized intersection in Thessaloniki, Greece, by classifying drivers into three categories: conservative, normal, and aggressive. The study used a logistic regression model to analyze the probability of stopping and crossing, on the basis of age, gender, vehicle speed, and the dilemma zone. A high percentage of drivers behaved aggressively, and drivers’ decisions were significantly correlate to their gender, vehicle’s speed, and the distance of the car from the stop line.

Based on video-camera data of drivers’ decisions at yellow lights, Elmitiny et al. [23] built classification-tree models to analyze the probability of stopping or crossing and of running a red light depending on traffic parameters.

Pulugurtha and Otturu [24] recently examined RLCs at 32 signalized intersections in the city of Charlotte, North Carolina, USA. They used the empirical bayes method to predict the number of crashes and then compared this predicted number with the actual number of crashes that occurred. They found that RLCs are effective in reducing the total number of crashes at signalized intersections.

Quiroga et al. [25], Yan et al. [26], and Tang et al. [27] investigated how various types of enforcement and traffic information systems (such as countdowns, flashing greens, and warning signs) affect driver behavior. To obtain a better understanding of driver decisions and behavior within the dilemma zone, majority of researchers have used the binary logistic regression model to analyze driver decisions during yellow-light intervals [28, 29, 30, 31]. Other researchers have modeled decisions with ordered-binary probit, binary probit, and fuzzy-logit models [32, 33, 34, 35]. Several studies have found that driver behavior in Riyadh tends to be aggressive [12, 13].

These local studies were published before implementation of STMES and in the absence of RLCs. The presence of an automated enforcement system at an intersection may make drivers less aggressive, as argued by Forrest et al. [36] and Shin and Washington [9].

Although automated enforcement systems at signalized intersections have been found to reduce the number of red-light violations by as much as three times[37,38], their effectiveness in terms of overall safety and their effect on decision-making at signalized intersections is still up for debate [39]. Research on the effects of automated enforcement systems on driver behavior and intersection safety in the context of Saudi Arabia is sparse.

3. METHODS

For assessing and analyzing driver behavior at signalized intersections, a binary logistic regression model is often suitable predicting the stopping probability based on multiple factors such as vehicle's speed and its location [28, 29, 30, 31]. The binary model can be represented as a follows [40, 41]:

�(� = ��) =�

� �, (1)



� = �� + �� ∗ ��

Where

� = probability of the driver choosing to stop;

� = binary decision variable (stop = 1 and cross = 0), dependent variable;

�� = constant coefficient;

�� = coefficients of independent variables;

�� = independent variables.

3.1. Dependent and Independent Variables

At signalized intersections, when the light changes to yellow, a driver must choose either to cross or to stop. This choice is a binary variable suitable for logistic modeling. The dependent variable is the driver’s decision regarding their behavior at a yellow signal, represented by �, where

� = 1 represents the choice to stop, and

� = 0 represents the choice to accelerate and cross the intersection.

The independent variables include two categorical and three continuous variables. Both the continuous and categorical variables were examined in the design of the model. The factors tested for inclusion in the model were as follows:

• approaching speed (speed);

• vehicle distance from the stop line at the onset of a yellow light (distance);

• installation of an enforcement system at the intersection (STMES: with = 1, without = 0);

• vehicle type (VEH: passenger car = 1, other = 0).

The variables selected for logistic modeling are described in Table 1.

A binary logistic regression model was developed to predict the probability of stopping based on these variables with and without STMES using the following equation:

�� (�) = �� + �� × �� + � × !"# + �$ × !%&'! + �( × )'*, (2)

Table 1 Variables Selected for Logistic Modeling

No. Variable Variable Description Codes/ Values Abbreviation

1 Vehicle Type Vehicle Type: Passenger

car or other

P. Car = 1

Other = 0 VEH

2 Enforcement

System Saher Traffic Monitoring

Enforcement System

With = 1

Without = 0 STMES

3 Distance (m) Vehicle’s distance from the stop line when the yellow

signal turns on

Continuous variable

Distance

4 Speed(km/h)

Vehicle’s speed when the yellow signal turns on

Continuous variable

Speed

3.2. Data Collection and Extraction

Data related to traffic flow and driver behavior was collected from 6:00am to 9: am on weekdays using a video camera, which was unobtrusively mounted in a building on the roadside 16–20 meters above the ground. Among the four signalized intersections which were chosen for the analysis, two of them were with STMES and the other two without STMES. A



total of 768 vehicles facing yellow signal and having to choice either to stop or cross were sampled. In good weather conditions, the data collected from the video recording included driver’s decision, signal status, traffic conditions, and vehicle characteristics. Reference points were marked at 5-m intervals to calculate the accurate position and speed of vehicles [30]. In this research, the reference points were divided into the following two categories to obtain more accurate measurements (Figure 1):

1. Major reference point. This represented the stop line, the Saher cameras (STMES), and the reference line (30 m from stop line).

2. Minor reference point. This represented the number of curbs on both sides (each curb was a fixed space of 0.5 m), especially in the absence of STMES at signalized intersections.

Figure 1 Location and distance of the reference points at the intersection

Video footage was studied frame-by-frame to extract the precise timings of signal changes and stop-line crossing or stopping events. The exact time was recorded as the signal turned yellow to a precision of the 30-Hz frame rate (1/30 second). At the same time, the vehicle closest to the stop line was identified as the target vehicle and its distance to the stop line was recorded. Virtual reference lines were superimposed on the video footage, along with the major and minor reference points, using U-lead software. Figure 2 shows a screenshot with the reference lines superimposed on the video.

Figure 2 Reference lines superimposed on the video



The distances of the vehicles from the stop line, vehicle speeds at the onset of a yellow signal, and driver decisions when the signal turned yellow were collected from the videos. Sample data extracted from the video are shown in Table 2.

Table 2 Sample of data extracted from the video

N

Time1 Driver action Distance

To Stop

Line At

Onset Of

Yellow

(m)

Speed

(m/s)

=

10/Dt

Speed

(km/hr)

Vehicle

type STMES

2 t1

(ss, ff)

t2

(ss,

ff)

Dt = t2-

t1

Second

Stop Cross

1 0.7 1.7 1 1 4 10.00 36.00 P.car with

2 16.1 17.2 1.1 1 6 9.09 32.73 P.car with

3 16.4 17 0.6 1 44 16.67 60.00 P.car with

4 14.3 15.1 0.8 1 6 12.50 45.00 P.car with

5 21.1 22.8 1.7 1 2 5.88 21.18 Other with

6 42.7 44.3 1.6 1 7 6.25 22.50 Other without

7 41.9 42.7 0.8 1 3 12.50 45.00 P.car without

8 40.3 41.5 1.2 1 2 8.33 30.00 P.car without

9 39.6 40.7 1.1 1 4 9.09 32.73 P.car without

10 56.7 57.5 0.8 1 6 12.50 45.00 other without

1. The time displayed on the counter consists of: ss: second, ff: fraction of second (1 ss = 30 ff).

2. STMES: Saher Traffic Monitoring Enforcement System

3.3. Estimated Logistics Model

The statistical results of the model and the probability of stopping (P) returned by the multivariate binary logistic regression model are listed in Table 3.

Table 3 Multivariate Binary Logistic Regression Model Statistics on Predictions (Stop/Cross)

Coefficients ++++ Std. Error z value Pr(>|z|) Odd Ratio

Intercept 1.4432 1.0338 1.396 0.163

Distance 0.1995 0.0419 4.765 <0.001 *** 1.22

Speed −0.1039 0.0272 −3.815 <0.001 *** 0.90

STMES 2.1897 0.6765 3.237 <0.01 ** 8.93

VEH −1.4421 0.8683 −1.661 0.097 0.24

Direction of the binary response variable ("0" indicates crossing / "1" indicates stopping)

Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

(Dispersion parameter for binomial family taken to be 1)

Null deviance: 193.62 on 139 degrees of freedom Residual deviance: 140.11 on 134 degrees of freedom

Most of the parameters considered in the model are logical and significant (at the 5% significance level). A relatively low level of confidence for vehicle type (VEH) may be attributed to a relatively lower number of observations. These predictions are expected to become significant with an increased number of observations of decision for heavy vehicles. Analysis of variance (ANOVA) results show that traffic factors, including distance (p < 0.001), speed (p < 0.001) and STMES (p < 0.001), have a significant effect on drivers’ decisions. This finding supports the results of other studies indicating that a driver’s decision



to stop or to cross is affected by both the approach speed and the distance from the stop line during the yellow phase at intersections[15, 19]..Results revealed that for the same vehicle location and velocity at the onset of yellow, the odds for stopping at a signalized intersection with STMES are 8.93 times those of a signalized intersection without STMES, and results in improving intersection safety by reducing red-light violations when STMES is present.

The probability curve for a vehicle stopping, as plotted in Figure 3, suggests that the average probability of a vehicle stopping is significantly increased when STMES is present and the vehicle is less than 10 meters from the stop line when the signal changes to yellow.

Figure 3 Probability of stopping during yellow light

Figure 3 shows the effect of distance to stop line on probability to stop for with and without STMES cases. With increasing distance to the stop line, the difference between the two curves quickly diminishes until the two curves merge. Vehicles are almost always less likely to stop when STMES is absent. In the yellow interval, when vehicles are at the same distance from the stop line, the average probability of stopping at intersections with STMES is significantly higher than without STMES by approximately 26%. This suggests that STMES is effective in deterring red-light violations, as evidenced by the large difference in stopping probability between these scenarios. As shown in Table 4, the model under-predicts crossing probability and aggregate prediction accuracy of the logistic model is 72.9 percent.

Table 4 Predictions of the logistic regression model

Predicted

Observed

Decision Percentage

Correct Cross Stop

Decision Cross 45 21 68.2

Stop 17 57 77.0

Aggregate Accuracy 72.9

3.4. Estimated Artificial Neural Network Model

In order to enhance model prediction accuracy, application of artificial neural network (ANN) is examined. ANN works like a collection of biological neurons. These neurons receive inputs from the source, perform a nonlinear operation, and predict an output [42]. An artificial neural



network takes the basic structure of the human brain [43], which is essentially very complex, nonlinear, and capable of parallel processing [44].

Recently, with developments in computer technology, artificial neural networks have been applied to many engineering problems such as for automotive applications [45, 46] and transportation planning [47]. Moreover, artificial neural networks have the ability to model all kinds of traffic violations and traffic patterns as long as an appropriate structure is implemented [48].

An artificial neural network comprises many simple and highly interconnected processing elements like the neurons in the human brain. The model can be defined as a black box comprising a series of equations for the calculation of outputs from given input values [49].

The block diagram of a neuron model is shown in Figure 4 and forms the basis for designing an artificial neural network [44]:

Figure 4 Block diagram of nonlinear model of a neuron

Mathematically, a neuron k can be described by the following equations:

,- = ∑ /-0102� 30, (3)

�- = 4(,- + 5-), (4)

Where

67= Bias (increasing or decreasing the net input of the activation function);

8�, 8 , … . 8< = inputs;

=7>�, =7>�, … , =7<= weights of the neuron k;

?7= linear combiner output due to input signals;

@(. )= activation function;

�7= output signal of the neuron.

Artificial neural networks use learning algorithms to simulate the function of the human brain [50]. A multilayer perceptron trained with a backpropagation algorithm is the simplest and the most widely used neural network method [47, 51]. The multilayer perceptron comprises three layers: input and output layers and one or more hidden layers.



A backpropagation algorithm was used as the learning algorithm for multilayered feedforward networks. The vehicle’s distance from the stop line when the yellow signal turns on (Distance), vehicle’s speed (Speed), presence of STMES, and vehicle type (VEH) were used as input parameters.

3.4.1. Artificial Neural Network results

The neural network package in the R software package was used to implement the artificial neural network model. The main structure of the artificial neural network is shown in Figure5.

Figure 5 Main structure of artificial neural network

We checked the implementation of neuralnet using the training and test data sets, and assessed model performance and prediction accuracy following the steps and pseudo code shown below:

Step 1: Read Data – Input variables (Speed, Distance, STMES, VEH) and Output (Decision).

Step 2: Model Calibration – Using Neuralnet package in R and saving output as ‘nnnn’ object.

Step 3: Check accuracy of neural network model with test data set.

R>nnnn$net.result[[1]]

R>nnn2 = ifelse(nnnn$net.result[[1]]>0.5,1,0)

R>misClasificationError = mean(DataFrame$Decision !=nnn2)

R>misClasificationError

[1] 0.1571428571

R> accuracy<- (1-misClasificationError)*100

R> accuracy

[1] 84.29



4. MODEL COMPARISON

The two models described above were compare on the basis of prediction accuracy. The artificial neural network model is clearly more accurate than the logistic regression model. The artificial neural network model made 11% more accurate predictions than the logistic regression model.

The estimated error values of both models show that the artificial neural network is more accurate than the binary logistic regression model for predicting the stopping probability at signalized intersections. The prediction accuracy for the logistic regression model was about 73 percent, whereas it was more than 84 percent for artificial neural network model. This might be caused by the ability of neural network model to account for the interaction effect among the input variables.

5. CONCLUSION

In the paper we compared the effectiveness of using binary logistic regression and artificial neural network models to assess traffic characteristics, safety implications, and performance at intersections with and without STMES. Major findings of the study are summarized below:

• The distance to the stop line is the most significant factor affecting drivers’ decision at the onset of yellow light.

• Vehicles are 26 percent more likely to stop at an intersection with STMES compered to intersection without STMES.

• The dilemma zone moved closer to the stop line, and its length decreased to from 35 to 10 m with installation of STMES.

• STMES increases the number of early stops as drivers try to avoid red-light violations. This finding supports findings reported elsewhere that the use of RLCs as automated enforcement devices increases the number of rear-end collisions (up to approximately 45 percent) [5, 52, 53, 54, 55].

• Comparison of two predictive modeling approaches show that artificial neural network is more accurate for predicting stopping behavior than binary logistic regression model. The artificial neural network model is 84.29 percent accurate, whereas binary logistic regression model accuracy is 72.86 percent.

The presence of STMES affects driver decisions about whether to slow down at a yellow signal thereby determines the boundary of dilemma zone, and in turn, affects the safety and performance at intersection. Increase in propensity to stop due to STMES induced perceived risk of penalty enhances perception of safety, and such systems are supported by 75 percent of road users in USA [56]. Predictive models can be used to enhance safety and assist drivers in decision making by dynamically updating the parameters of intersection operations and adjusting the yellow-light period to reduce the effect of the dilemma zone at signalized intersections. Further research is required to integrate predictive models with real-time signal operations and advanced warning systems. Future research with more field data could further improve predictive accuracy of the neural network model.

REFERENCES

[1] National Highway Traffic Safety Administration. HTSA’s Fatality Analysis Reporting System (FARS) 2008: http://www.fars.nhtsa.dot.gov/Main/ index.aspx

[2] Statistical Report. General Department of Traffic, Monitory of Interior, 2009, available at: http://www.moi.gov.sa/ (Accessed: 5 February 2015).



[3] Lawrence, E.D., Thomas, L., Srinivasan, R. and Staplin, L. Automated Enforcement: A Compendium of Worldwide Evaluations of Results, National Highway Traffic Safety Administration, Report No. DOT HS 810 763, 2007, USA.

[4] Retting R.A.; Williams A.F.; Preusser D.F.; Weinstein H.B. Classifying urban crashes for countermeasure development. Accid Anal Prev 1995, 27, 283–294, DOI: 10.1016/0001-4575(94)00068-W.

[5] De Pauw E.; Daniels S.; Brijs T.; Hermans E.; Wets G. To brake or to accelerate? Safety effects of combined speed and red light cameras. J. Safety Research 2014, 50, 59–65, DOI: 10.1016/j.jsr.2014.03.011.

[6] Martinez K.L.H.; Porter B.E. Characterizing red light runners following implementation of a photo enforcement program. Accid Anal Prev 2006, 38, 862–870, DOI: 10.1016/j.aap.2006.02.011. 2006.

[7] Llau A.F.; Ahmed N.U. The effectiveness of red light cameras in the United States–a literature reviews. Traffic Inj Prev 2014, 15, 542–550, DOI: 10.1080/15389588.2013.845751.

[8] Garber N.C.; Miller J.S.; Abel R.E.; Eslambolchi S.; Korukonda S. The impact of red light cameras (photo-red enforcement) on crashes in Virginia. FHWA/VTRC/07R2. Virginia Transportation Research Council. Charlottesville, Virginia, 2007.

[9] Shin K.; Washington S. The impact of red light cameras on safety in Arizona. Accid Anal Prev 2007, 39, 1212–1221, DOI: 10.1016/j.aap.2007.03.010.

[10] Retting R.A.; Ferguson S.A.; Hakkert A.S. Effects of Red Light Cameras on Violations and Crashes: A Review of the International Literature. Traffic Inj Prev 2003, 4, 17–23, DOI: 10.1080/15389580309858

[11] Fitzsimmons E.J.; Hallmark S.L.; Orellana M.; McDonald T.; Matulac, D. Investigation of violation reduction at intersection approaches with automated red light running enforcement cameras in Clive, Iowa, using a cross-sectional analysis. J Transp Eng 2009, 135, 984–989, DOI: 10.1061/%28ASCE%29TE.1943-5436.0000079.

[12] Al-Ghamdi, A.S. Entering Headway for Through Movements at Urban Signalized Intersections, Transportation Research Board, 78th Annual Meeting, Washington D.C., 1999, pp. 42-47.

[13] Al-Ghamdi A.S. Analysis of Time Headway on Urban Road: Case Study from Riyadh. J Transp Eng 2001, 127, 289–294, DOI: 10.1061/(ASCE)0733-947X(2001)127:4(289).

[14] Alam J.B.; Osra K.; Al-Bar H.; Shaher Z. Signalized Intersection Capacity Adjustment Factors for Makkah, Saudi Arabia. Canadian Journal of Transportation 2011, 4.

[15] Yang Z.; Tian X.; Wang W.; Zhou X.; Hongmei L.H. Research on driver behavior in yellow interval at signalized intersections, Hindawi Publishing Corporation. Math Probl Eng 2014, 518782, DOI: 10.1155/2014/518782.

[16] Petridou E.; Moustaki M. Human Factors in the Causation of Road Traffic Crashes. Eur J Epidemiol 2000, 16, 819-826, DOI: 10.1023/A: 100764980.

[17] Ellison A.B.; Greaves S. Speeding in urban environments: Are the time savings worth the risk? Working Paper ITLS-WP-15-10, 2015. Institute of Transport and Logistics Studies (C13), The Australian Key Centre in Transport and Logistics Management, The University of Sydney.

[18] Martinez R. Testimony before the House Committee on Transportation and Infrastructure Subcommittee on Surface Transportation, 1997. (http://house.gov/transportation/surface/sthearin/ ist717/martinez.htm)

[19] Lum K.M.; Wong Y.D. A before-and-after study of driver stopping propensity at red light camera intersections. Accid Anal Prev 2001, 35, 111–120, DOI: 10.1016/S0001-4575(01)00096-3.

[20] Lum K.M.; Wong, Y.D. A study of stopping propensity at matured red light camera T-intersections. J Safety Res 2002, 33, 355–369, DOI: 10.1016/S0022-4375(02)00031-2.



[21] Lum K.M.; Wong Y.D. Effects of red light camera installation on driver behavior at a signalized cross-junction in Singapore. Road and Transport Research 2002, Vol 11 No 3. 2002b

[22] Papaioannou P. Driver behavior, dilemma zone and safety effects at urban signalized intersections in Greece. Accid Anal Prev 2007, 39, 147–158, DOI: 10.1016/j.aap.2006.06.014.

[23] Elmitiny N.; Yan X.; Radwan E.; Russo C.; Nashar D. Classification analysis of driver’s stop/go decision and red light running violation. Accid Anal Prev 2010, 42, 101–111, DOI: 10.1016/j.aap.2009.07.007.

[24] Pulugurtha S.S.; Otturu R. Effectiveness of red light running camera enforcement program in reducing crashes: evaluation using before the installation, after the installation, and after the termination data. Accid Anal Prev 2014, 64, 9–17, DOI: 10.1016/j.aap.2013.10.035.

[25] Quiroga C.; Kraus E.; Schalkwyk I.V.; Bonneson J. Red light running–A policy review, Texas Transportation Institute, College Station, TX, 2003

[26] Yan X.; Radwan E.; Guo D.; Richards S. Impact of signal ahead pavement marking on driver behavior at signalized intersections. Transp Res F Traffic Psychol Behav 2009, 12, 50–67, DOI: 10.1016/j.trf.2008.07.002.

[27] Tang K.; Xu Y.; Wang P.; Wang F. Impacts of Flashing Green on Dilemma Zone Behavior at High-Speed Intersections: Empirical Study in China. J Transp Eng 2015, 141, 04015005, DOI: 10.1061/(ASCE)TE.1943-5436.0000770.

[28] Gates T.J.; Noyce D.A.; Laracuente L. Analysis of dilemma zone driver behavior at signalized intersections, Transportation Research Board, Washington D.C., 2007.

[29] Li Z. Dynamic dilemma zone modeling and its protection．PhD Thesis，University of

Cincinnati，Cincinnati, USA, 2011

[30] Long K, Liu Y, Han LD. Impact of countdown timer on driving maneuvers after the yellow onset at signalized intersections: An empirical study in Changsha, China. Safety science. 2013 Apr 30; 54:8-16.

[31] Zhang Y.; Fu C.; Hu L. Yellow light dilemma zone researches: a review. J Traffic Transp Eng Engl Ed 2014, 1, 338–352, DOI: 10.1016/S2095-7564(15)30280-4.

[32] Li P.; Abbas M. Stochastic dilemma hazard model at high- speed signalized intersections.

J Transp Eng 2010, 136, 448–456, DOI: 10.1061/(ASCE)TE.1943-5436.0000066．

[33] Sharma, A.; Bullock, D.; Peeta, S. Estimating dilemma zone hazard function at high speed isolated intersection. Transp Res Part C Emerg Technol 2011, 19, 400–412, DOI:

10.1016/j.trc.2010.05.002．

[34] Liu Y.; Chang G.L.; Yu J. Empirical study of driver responses during the yellow signal phase at six Maryland intersections. J Transp Eng 2012, 138, 31–42, DOI: 10.1061/(ASCE)TE.1943-5436.0000278

[35] Moore, D.; Hurwitz, D.S. Fuzzy logic for improved dilemma zone identification: driving simulator study. Transp Res Rec 2013, 2384, 25–34, DOI: 10.3141/2384-04.

[36] Forrest C.M.; Srinivasan R.; Masten S.; Carte D.; Reurings M. Development of a Speeding-Related Crash Typology: Summary Report. Federal Highway Administration. FHWA-HRT-10-039. 2005.

[37] Retting R.A.; Williams A.F.; Farmer C.M.; Feldman A.F. Evaluation of red light camera enforcement in Fairfax, Virginia, USA. ITE J 1999, 69, 30–34, DOI: 10.1016/S0001-4575(98)00059-1.

[38] Retting R.A.; Williams A.F.; Farmer C.M.; Feldman A.F. Evaluation of red light camera enforcement in Oxnard, California. Accid Anal Prev 1999, 31, 169–174, DOI: 10.1016/S0001-4575(98)00059-1.

[39] McGee H.W. Making intersections safer: a toolbox of engineering countermeasures to reduce red-light running: an informational report. Inst of Transportation Engineers. 2003.



[40] Feinberg, S., 1980. The Analysis of Cross-Classified Categorical Data. MIT Press, Cambridge, MA.

[41] Hosmer, D.W. and Lemeshow, S., 1989. Applied Logistic Regression. Wiley, New York.

[42] Yusaf T.F.; Yousif B.F.; Elawad M.M. Crude palm oil fuel for diesel-engines: experimental and ANN simulation approaches. Energy 2011, 36, 4871–4878, DOI: 10.1016/j.energy.2011.05.032.

[43] Ata R. Artificial neural networks applications in wind energy systems: a review. Renew Sust Energ Rev 2015, 49, 534–562, DOI: 10.1016/j.rser.2015.04.166.

[44] Haykin, S. Neural networks: a comprehensive foundation. Prentice Hall, Ontario, Canada, 1999.

[45] Aydogan H. Prediction of diesel engine performance, emissions and cylinder pressure obtained using Bioethanol-biodiesel-diesel fuel blends through an artificial neural network. JESA 2015, 26, 74–83.

[46] Tosun E.; Aydin K.; Bilgili M. Comparison of linear regression and artificial neural network model of a diesel engine fueled with biodiesel-alcohol mixtures. Alexandria Engineering Journal 2016, 55, 3081–3089, DOI: 10.1016/j.aej.2016.08.011.

[47] Ahmed A.; Gazder U. Traffic Forecasting for King Fahd Causeway: Comparison of Parametric Technique with Artificial Neural Networks. In Proceeding of the International Conference on Advances in Structural, Civil and Environmental Engineering SCEE, 2013, Vol. 2013, pp. 75–79.

[48] Qiao F.; Yang H.; Lam W. Intelligent simulation and prediction of traffic flow dispersion. Transport Res B Meth 2001, 35, 843–863, DOI: 10.1016/S0191-2615(00)00024-2.

[49] Ozgoren M.; Bilgili M.; Sahin B. Estimation of global solar radiation using ANN over Turkey, Expert Syst Appl 2012, 39, 5043–5051, DOI: 10.1016/j.eswa.2011.11.036.

[50] Turban, E., & Aronson, J. E. D. S. S. (2001). Intelligent Systems. Pretince Hall. New Jersey.

[51] Pradhan B.; Lee S. Landslide susceptibility assessment and factor effect analysis: backpropagation artificial neural networks and their comparison with frequency ratio and bivariate logistic regression modelling. Environ Model Soft 2010, 25, 747–759, DOI: 10.1016/j.envsoft.2009.10.016.

[52] Erke A. Red light for red-light cameras? A meta-analysis of the effects of red light cameras on crashes. Accid Anal Prev 2009, 41, 897–905, DOI: 10.1016/j.aap.2008.08.011.

[53] Høye A. Still red light for red light cameras? An update. Accid Anal Prev 2013, 55, 77–89, DOI: 10.1016/j.aap.2013.02.017.

[54] Vanlaar W.; Robertson R.; Marcoux K. An evaluation of Winnipeg’s photo enforcement safety program: results of time series analyses and an intersection camera experiment. Accid Anal Prev 2015, 62, 238–247, DOI: 10.1016/j.aap.2013.09.023.

[55] Polders E.; Cornu J.; Ceunynck T.D.; Daniels S.; Brijs K.; Brijs T.; Hermans E.; Wets G. Drivers’ behavioral responses to combined speed and red light cameras. Accid Anal Prev 2015, 81, 153–166, DOI: 10.1016/j.aap.2015.05.006.

[56] Retting R.A.; Williams A.F. Red light cameras and the perceived risk of being ticketed. Traffic Eng Control 2000, 41,224– 227.

effects of automated traffic enforcement on driver …€¦ · saad al jaman* department of civil...

Documents