modelling_and_simulation_of_transportation_systems_a_scenario_planning_approach-libre (1).pdf

8/9/2019 Modelling_and_Simulation_of_Transportation_Systems_a_Scenario_Planning_Approach-libre (1).pdf

1/12

ISSN 0005—1144ATKAAF 50(1—2), 39—50 (2009)

George Papageorgiou, Pantelis Damianou, Andreas Pitsillides, Thrasos Aphamis, Demetris Charalambous, Petros Ioannou

Modelling and Simulation of Transportation Systems:

a Scenario Planning Approach

UDK 656.132.021.24IFAC 5.7.1; 2.8.1

Original scientific paper

It is increasingly realized that building more roads does not solve the traffic congestion problem but actu-ally makes it worse. Instead of adding capacity to our roads there should be an effort to find ways in order to enhance the level of service of the public transport mode especially with the use of technology, such asapplying computer and information technology to transportation systems. Advanced technologies such as

Intelligent Transportation Systems (ITS) provide a big opportunity for alleviating the traffic congestion prob-lem. On the other hand, ITS technologies require though rigorous testing and evaluation, which can only beachieved with computer simulation modeling. This paper presents an overview on traffic simulation as wellas the development process of a microscopic simulation model of a highly congested traffic network in Nicosia, Cyprus. The validated simulation model gives transportation planners and traffic managers the capa- bility to test various Bus Rapid Transit scenario solutions involving the use of intelligent transportation sys-tems prior to their implementation.

Key words: Modeling, Simulation, Transportation Systems, Bus Rapid Transit.

1 INTRODUCTION

Traffic congestion constitutes a complex dynam-ical problem. It comprises many complex process-es, and incorporates many elements interacting witheach other such as vehicles, driver behaviors, roadgeometry, traffic signs and so on. In such a com-

plex problem situation a computer simulation approach can be very effective by providing evalua-tions for various traffic conditions. It can help pol-icy-makers understand and analyze traffic, assesscurrent problems and propose plausible solutions.Traffic simulation can support transportation plan-ning, and traffic management decision making, for

a long-term sustainable urban development. Thenew solutions and techniques can effectively betested in a »virtual reality« environment, in thecomfort of one’s office, without disrupting the roadtraffic or having to leave for field trials. Effectivetraffic management and control strategies though,require up-to-date valid simulation test results.

This paper presents the current progress and les-sons learned from the modeling and simulation of an urban traffic network. The work presented inthis paper is part of the Trafbus research project,

partially funded by the Cyprus Research Promotion

Foundation. The Trafbus research project is con-

cerned with the modeling, simulation and analysisof traffic flow for a major traffic network in

Nicosia, Cyprus. Further, the use of Bus RapidTransit Systems are to be evaluated and tested in asimulated environment for increasing the level of service of the bus transport mode and optimizingtraffic flow.

In particular, a microscopic simulation model for Strovolos Avenue, which is depicted in Figure 1,is developed, which is the main transport artery of the Strovolos Municipality. Strovolos exhibits thehighest traffic flows as compared with the other re-gions. Further, Strovolos Avenue serves as the con-nector between Nicosia and a large and heavily

populated area of urban and rural communities in-cluding Strovolos which is the largest municipalarea, Lakatameia, Tseri, Deftera and others. Basedon the simulation model of Strovolos Avenue pre-sented in this paper various scenarios of microscop-ic models of traffic flow including »dedicated buslanes« and Bus Rapid Transit (BRT) Systems, areto be studied using computer simulation for the

purpose of designing a more effective and safer traffic network. The Trafbus research project rep-resents a pioneering work for Cyprus, and the ex-

periences gained from the modeling and simulation

of Strovolos Avenue will serve as a knowledge

AUTOMATIKA 50(2009) 1—2, 39—50 39


2/12

veloped to characterize the relationships among thetraffic stream variables of speed, flow, and concen-tration or density. Human factor modeling [7], dealswith salient performance aspects of the human ele-ment in the context of the human-machine interac-tive system. These include perception-reactiontime, control movement time, responses to: trafficcontrol devices, movement of other vehicles, haz-ards in the roadway, and how different segmentsof the population differ in performance. Further,human factors theory deals with the kind of con-trol performance that underlies steering, braking,and speed control. Human factors theory providesthe basis for the development of Car followingmodels. Car following models [8], examine themanner in which individual vehicles (and their drivers) follow one another. In general, they are

developed from a stimulus-response relationship,where the response of successive drivers in the traf-fic stream is to accelerate or decelerate in propor-tion to the magnitude of the stimulus. Car follow-ing models recognize that traffic is made up of dis-crete particles or driver-vehicle units and it is theinteractions between these units that determinedriver behaviour, which affects speed-flow-density

patterns. On the other hand, continuum models [9]are concerned more with the overall statistical be-haviour of the traffic stream rather than with theinteractions between the particles. Following thecontinuum model paradigm, macroscopic flowmodels [10], discard the microscopic view of traf-fic in terms of individual vehicles or individual sys-tem components (such as links or intersections) andadopt instead a macroscopic view of traffic in anetwork. Macroscopic Flow Models consider vari-ables such as flow rate, speed of flow, density andignore individual responses of vehicles. Traffic im-

pact models [11] deal with traffic and safety, fuelconsumption and air quality models. Traffic andsafety models describe the relationship betweentraffic flow and accident frequency. Unsignalizedintersection theory [12] deals with the gap accept-

ance theory and the headway distributions used ingap acceptance calculations. Traffic flow at signal-ized intersections [13] deals with the statistical the-ory of traffic flow, in order to provide estimates of delays and queues at isolated intersections, includ-ing the effect of upstream traffic signals. Trafficsimulation modeling [14] deals with the trafficmodels that are embedded in simulation packagesand the procedures that are being used for conduct-ing simulation experiments.

In a two-level view the problem of modeling ve-hicle traffic flow, may be approached mathemati-

cally by two main scales of observation: the mi-croscopic and the macroscopic levels. In the mi-

40 AUTOMATIKA 50(2009) 1—2, 39—50

Modelling and Simulation of Transportation Systems: A Scenario Planning... G. Papageorgiou, et al.

base for similar projects to be carried out all over Cyprus.

2 TRAFFIC FLOW MODELING

The study of traffic flow [1], and in particular vehicular traffic flow, is carried out with the aimof understanding and assisting in the preventionand remedy of traffic congestion problems. Thefirst attempts to develop a mathematical theory for traffic flow date back to the 1930s [2, 3], but evenuntil today we do not have a satisfactory and gen-eral mathematical theory to describe real trafficflow conditions. This is because traffic phenomenaare complex and nonlinear, depending on the inter-actions of a large number of vehicles. Moreover,vehicles do not interact simply by following thelaws of physics, but are also influenced by the psy-chological reactions of human drivers. As a resultwe observe chaotic phenomena such as cluster for-mation and backward propagating shockwaves of

vehicle speed/density [4] that are difficult, if even possible, to be accurately described with mathe-matical models. According to a state of the art re-

port of the Transportation Research Board [5],mathematical models for traffic flow may be clas-sified as: Traffic Stream Characteristics Models,Human Factor Models, Car Following Models,Continuum Flow Models, Macroscopic FlowModels, Traffic Impact Models, Unsignalized Inter-section Models, Signalized Intersection Models andTraffic Simulation Models. Below we describe

briefly each of the above categories.

Traffic stream characteristics [6] theory involvesvarious mathematical models, which have been de-

Fig. 1 The Strovolos Avenue Traffic Network


3/12

AUTOMATIKA 50(2009) 1—2, 39—50 41

G. Papageorgiou, et al. Modelling and Simulation of Transportation Systems: A Scenario Planning...

croscopic level every vehicle is considered as anindividual, and therefore for every vehicle we havean equation, that is usually an ordinary differentialequation. In a macroscopic level we use the analo-gy with fluid dynamics models, where we have asystem of partial differential equations which in-volves variables such density, speed, flow rate of traffic stream with respect to time and space. Insuch a configuration speed, flow, and density will

be governed by the following partial differentialequation [9].

(1)

where q is the traffic flow, x is the displacement, ρ is the density and g ( x, t ) is function for sourcesand sinks of traffic flow.

The microscopic model involves separate unitswith characteristics such as speed, acceleration andindividual driver-vehicle interaction. Microscopicmodels may be classified in different types basedon the so called car-following modelling approach.The car-following modelling approach implies thatthe driver adjusts his or her acceleration accordingto the conditions of leading vehicles. In these mod-els the vehicle position is treated as a continuousfunction and each vehicle is governed by an ordi-nary differential equation that depends on speedand distance of the car in the front as shown in Eq.(2).

(2)

Where x f is the one-dimensional position of thefollowing vehicle, xl is the one-dimensional posi-tion of the leading vehicle, t is time, and λ is asensitivity coefficient.

Another type of microscopic models are the Cel-lular Automata or vehicle-hopping models whichdiffer from the car-following approach in that they

are fully discrete time models. They consider theroad as a string of cells which are either empty or occupied by one vehicle. One such model is theStochastic Traffic Cellular Automata [15, 16].

Microscopic approaches are generally computa-tionally intense, as each car has an ODE to besolved at each time step, and as the number of carsincreases, so does the size of the system to besolved. Analytical mathematical microscopic mod-els are difficult to evaluate but a remedy for thisis the use of microscopic computer simulation. Insuch microscopic traffic models, vehicles are treat-

ed as discrete driver-vehicle units moving in acomputer-simulated environment.

&&

& &

x t x t x t

x t x t f

l f

l f

( )( ) ( )

( ) ( )= ⋅

−

− λ

∂

∂+

∂

∂=

q

x t g x t

ρ( , )

On the other hand, macroscopic models aim atstudying traffic flow using a continuum approach,where it is assumed that the movement of individ-ual vehicles exhibit many of the attributes of fluidmotion. As a result, vehicle dynamics are treatedas fluid dynamics. This idea provides an advantagesince detailed interactions are overlooked, and themodel's characteristics are shifted toward the moreimportant parameters such as flow rate, concentra-tion or traffic density, and average speed, all beingfunctions of one-dimensional space and time. Thisclass of models is represented by partial differen-tial equations. Modeling vehicular traffic via mac-roscopic models is achieved using fluid flow theo-ry in a continuum responding to local or non-localinfluences. The mathematical details of such mod-els are less than those of the microscopic ones. The

drawback of macroscopic modeling is the assump-tion that traffic flow behaves like fluid flow whichis a rather harsh approximation of reality. Vehiclestend to interact among themselves and are sensi-tive to local traffic disturbances, phenomena whichare not captured by macroscopic models. On theother hand, macroscopic models are suitable for studying large scale problems and are computation-ally less intense especially after approximating the

partial differential equation with a discrete time fi-nite order equation.

There exists also a third level of analysis the socalled mesoscopic level, which is somewhere be-tween the microscopic and the macroscopic levels.In a mesoscopic or kinetic scale, which is an inter-mediate level, we define a function f (t , x, v) whichexpresses the probability of having a vehicle attime t in position x at velocity v. This function,following methods of statistical mechanics, can becomputed solving an integro-differential equation,like the Boltzmann Equation [17, 18].

The choice of the appropriate model depends onthe level of detail required and the computing

power available. As a result of advancements incomputer technology in recent years, the trendtoday is towards utilising microscopic scale mathe-matical models, which incorporate human factorsand car-following models as a driver-vehicle be-haviour unit.

3 MICROSCOPIC TRAFFIC MODELING

SOFTWARE

Microscopic simulation is a term used in trafficmodelling and is typified by software packages

such as VISSIM [19, 20, 21, 22], CORSIM [23,24, 25, 26], AIMSUN [27, 28, 29, 30] and PARA-


4/12

MICS [31, 32, 33, 34]. Traffic simulation micro-scopic models simulate the behaviour of individualvehicles within a predefined road network and areused to predict the likely impact of changes in traf-

fic patterns resulting from proposed commercial de-velopments or road schemes. They are aiming tofacilitate transportation consultants municipalities,government transportation authorities and publictransportation companies. The traffic flow modelsused are discrete, stochastic, time step based mi-croscopic models, with driver-vehicle units as sin-gle entities.

Traffic simulation software modelers combine ina single package multiple traffic flow mathemati-cal models and therefore make it possible to com-

bine the current knowledge on traffic theory when

analyzing a traffic congestion problem.A screenshot of the VISSIM graphical user in-

terface is provided in Figure 2. The microscopicmodel depicted in the figure was developed inorder to analyze traffic and evaluate the impact of various bus priority scenarios for a traffic network in Nicosia, Cyprus [35, 36].

In today’s traffic simulation software, data suchas network definition of roads and tracks, techni-cal vehicle and behavioural driver specifications,car volumes and paths can be inserted in a graphi-cal user interface mode. Values for accelerationmaximum speed and desired speed distributions can

be configured by the user to reflect local trafficconditions. Various vehicles types can also be de-fined. Further traffic control strategies and algo-rithms may be defined as well as interfaces may

be build with well known urban traffic controllers.Below we present some of their main featuresCORSIM, PARAMICS, VISSIM, and AIMSUN,which were calibrated and validated in a number

of traffic studies worldwide.CORSIM, which stands for Corridor Microscop-

ic Simulation is developed by Federal HighwayAdministration of United States. It has evolvedfrom two separate traffic simulation programs NET-SIM and FRESIM. NETSIM models arterials withsignalised and unsignalised intersections, whileFRESIM models uninterrupted freeways and urbanhighways.

In the case of VISSIM developed by PTV, themicroscopic model consists of a psycho-physical

car following model for longitudinal vehicle move-ment and a rule-based lane changing algorithm for lateral movements. The model is based on an urbanand a freeway model which were developed byWiedemann from the University of Karlsruhe. VIS-SIM is especially renowned for its signal controlmodule, which by using a vehicle actuated pro-gramming language can model almost any trafficcontrol logic. Further, VISSIM scores high on itsability to model public transportation systems.

AIMSUN was developed by TSS in order tosimulate urban and interurban traffic networks. It

is based on the car-following model of Gibbs.AIMSUN therefore is based on a collision avoid-ance car-following model. Traffic can be modelledvia input flows and turning movements origin des-tination matrices and route choice models.

42 AUTOMATIKA 50(2009) 1—2, 39—50


Fig. 2 The VISSIM Graphical user Interface Depicting Part of Strovolos Ave


5/12

AUTOMATIKA 50(2009) 1—2, 39—50 43


PARAMICS which stands for Parallel Micro-scopic Simulation comprises of various moduleswhich include a modeller, a processor, an analyser,a monitor, a converter and an estimator. PARA-MICS is renowned for its visualization graphicsand for its ability to model quite a diverse rangeof traffic scenarios.

A comprehensive review of simulation modelsof traffic flow was conducted by the Institute for Transport Studies at the University of Leeds as partof the SMARTEST Project, a collaborative projectto develop micro-simulation tools to help solveroad traffic management problems [37]. The studycompared the capabilities of more than 50 simula-tion packages. The results are available on theinternet at http://www.its.leeds.ac.uk/projects/

/smartest. Other significant reviews of traffic simu-lation software include the work of Bloomberg andDale [38] who compared Corsim and Vissim aswell as the work of Boxill and Yu [39] who com-

pared the capabilities of Corsim, Aimsun andParamics. It can be concluded from the various re-views that software modelers, which have compar-ative capabilities include VISSIM, AIMSUN, andPARAMICS.

In a more recent comparative study of micro-scopic car-following behaviour Panwai and Dia[40] evaluate AIMSUN, VISSIM and PARAMICS.

They conclude that the accuracy of a traffic simu-lation system depends highly on the quality of itstraffic flow model at its core, which consists of car following and lane-changing models. In the studythe car-following behaviour for each simulator wascompared to field data obtained from instrumentedvehicles travelling on an urban road in Germany.The Error Metric on distance [41] performance in-dicator gave substantially better values for AIM-SUN than those of VISSIM and PARAMICS.Further the Root Mean Square Error (RMSE) wassubstantially less for VISSIM and AIMSUN than

the RMSE for PARAMICS. In another paper pre-sented at the 9th TRB Conference on the Applicat-ion of Transportation Planning Methods Choa et al.[42] concluded that although CORSIM provides theshortest traffic network setup time PARAMICS andVISSIM generated simulation results that better matched field observed conditions and traffic engi-neering principles.

4 THE PROBLEM OF TRAFFIC CONGESTION

IN CYPRUS

In 2007 there were more than 600000 registeredvehicles in Cyprus, a figure that approaches the

number of people of the island and unfortunately puts Cyprus at the top of the world when it comesto number of vehicles per capita. As a result, trav-el times are increasing and traffic congestion be-comes an everyday reality. On a daily basis we areconfronted with rush hours, road accidents, air pol-lution and driver-stress, causing an increasing num-

ber of economic, social and environmental prob-lems.

Some of the causes of the current traffic conges-tion situation include the rapid economic develop-ment in Cyprus as well as the concentration of pop-ulation in urban communities. Further, people areturning away from using the bus and use their own

private car for daily transportation. As a result,Cyprus cities have serious traffic congestion prob-

lems in main arterials such as Strovolos Avenue,and at signalized intersections.

According to a recent traffic survey carried out by the local newspaper »Politis« the average resi-dent of Nicosia (the Capital of Cyprus) drives for 1 hour and 40 minutes per day for relatively smalldistances. Further, 76 % of the survey respondentsdeclare they have never used the bus. The promis-ing finding was that 57 % declare they would usethe bus if its quality of service was better.

Even though the Public Works Department of theMinistry of Communications and Works buildsmore and more roads in order to meet capacity de-mands the traffic congestion problem becomesworse. And the reason is that we use more andmore cars for our everyday transportation. In thelast three decades there was an unprecedented in-crease in the number of vehicles. The number of registered vehicles has increased from 100000 in1980 to more than 600000 by the year 2007, whichrepresent more than a 600 % increase in 27 years.

In addition to the increased number of cars, theuse of the bus transport mode has sharply fallen.

From 13 million passengers during the year 1981we are down to 3 million for the year 2007. Thisrepresents more than a 400 % decrease in the useof the bus transport mode.

The traffic congestion problem situation will re-main and even get worse, unless the traffic flowtrends and needs are understood and analyzed for deriving effective solutions. Traffic congestion con-stitutes a highly complex dynamical problem,which consists of a combination of factors such asthe lack of a modern public transport system, thedependence on the private car, the town structure

and the urban environment, the radial road systemand the incomplete primary road infrastructure.


6/12

The assumption held by most policy makers, city planners, and transportation officials is that trafficvolume is exogenous in that traffic volume is grow-ing as population grows and local economy devel-ops. Building roads therefore should keep traveltime at low levels but also can serve special inter-ests of particular businesses which would be eager to satisfy the personal interest of the policy maker.However traffic volume is not exogenous, that isroad building does not alleviate traffic congestion.The number of cars in a particular region is a major determinant of traffic volume. Total traffic volumetherefore equals the number of vehicles in the re-gion multiplied by the distance traveled of each ve-hicle per day. In turn the distance traveled per dayfor each vehicle is equal to the number of tripsmultiplied by the length of each trip. The number

of trips per day and the average trip length are notconstant but depend on the level of traffic conges-tion. People will take additional trips if the trafficis light while they would stick to the necessarytrips when the traffic is heavy. Further, the number of vehicles in the region is not constant but varieswith respect to the population multiplied by thenumber of cars per person. Furthermore the num-

ber of vehicles per person or business is not con-stant but depends on the attractiveness of driving,which depends on the level of traffic congestion.

Realizing that by building more roads the traffic

congestion problem would even get worse, theGovernment of Cyprus and particularly theMinistry of Communications and Works aim for amore modern transport policy. The policy involvesrestraining the use of private cars, the enhancementof the urban bus transport system and bettermentof its level of service, the promotion of alternativemeans of transport such as the bicycle, and the con-struction of a modern urban road network.

There is a need for advanced mathematical meth-ods and models in order to analyze the dynamicityand chaotic behavior involved in the traffic con-gestion problem situation. The following sectiondescribes the development of traffic flow theories,which aim to analyze the traffic congestion prob-lem.

5 THE PROPOSED TRAFFIC MODELING AND

SIMULATION METHOD

As described in the previous sections traffic phe-nomena constitute a chaotic dynamical problem sit-uation, which make traffic modelling and simula-tion a very complex, iterative and difficult process.In order to increase our chances for a successfulsimulation model the following methodology is

proposed, which is applied for modeling the Stro-volos Avenue traffic network as described in thenext section.

The proposed traffic modeling and simulationmethod is based on the suggestions of Liebermanand Rathi [14]. As shown in Figure 3, the first stepis to identify and clearly define the problem.Caution should be taken not to just deal with thesymptoms of the problem but to find the real causeof the traffic congestion problem. The use of ad-vanced modeling techniques such as system dy-namics analysis may be incorporated here. Having

defined the problem, together with specifying the problem's extent and significance, the next step isto define the model objectives. In other words the

purpose of the simulation model to be developedshould be specified. Again here caution should betaken to investigate whether the stated objectiveswould really solve the problem as well as to takeinto consideration any »side effects« that mayoccur as a result of specific proposed solutions.

44 AUTOMATIKA 50(2009) 1—2, 39—50


Fig. 3 The Proposed Traffic Modeling and Simulation Method


7/12

AUTOMATIKA 50(2009) 1—2, 39—50 45


Next, we have the definition of the system to bestudied. Here the major components of the systemneed to be identified as well as the boundary of the domain of the system needs to be defined.

Further the interactions of the various componentsneed to be analyzed. Finally, the necessary infor-mation to be used as input to the model has to beidentified. Once the system to be studied is definedthe next step would be the development of the mo-del. Here, the level of complexity needed to satis-fy the stated objectives should be identified. Itmight be for example that a macroscopic modelwould be adequate for the current problem. There-fore the model is to be classified and its inputs andoutputs should be defined.

During model development an appropriate soft-

ware modeler based on the model objectives should be selected. It is important therefore to carry outan assessment of a number of software for trafficsimulation and investigate there capabilities andlimitations. In case the selected model is not ade-quate for the model objectives then enhancementsshould be carried out with some further softwaredevelopment. In this event the flow of data withinthe model as well as some functions and processesof the model components need to be defined.Further, the calibration requirements need to be de-termined. Mathematical, logical and statistical al-

gorithms of each inadequately represented (by theselected software modeller) system component withits activities and interactions should be developed.The logical structure for integrating these modelcomponents needs to be created to support the flowof data among them. Then the software develop-ment method should be selected as well as an ap-

propriate programming language, user interface andthe presentation format of model results. Finally,the design logic and all computational proceduresshould be documented and the software codeshould be developed and debugged.

The next step which partly belongs to the modeldevelopment process is model calibration. Here thenecessary data should be collected or acquired inorder to calibrate the model. Then, this data should

be introduced into the model. Further, especiallyfor the case of software development enhancementsit should be verified that the software executes inaccordance with the design specifications. Modeldevelopment and calibration are part of an itera-tive process as shown in Figure 3 that leads to val-idating the model. Model validation includes thecollection or acquisition of data as well as reduc-

tion and organizing for purposes of validation. Thatis, gathered data should be reduced and structured

in such a way so that they are in the same formatas the data generated by the model. Further, vali-dation criteria should be established stating the un-derlying hypotheses and selecting the statistical

tests to be applied.The iterative process of model development, cal-

ibration, verification and validation is completedonce it is established that the model describes thereal system at an acceptable level of accuracy over its entire domain of operation. Here, the use of sta-tistical testing methods proves to be useful. In moredetail the validation process may include experi-mental design development, identification of thecauses for any failure to satisfy the validation testsand repairing the model accordingly. As differences

between the model results and real world data

emerge the developer must repair the model andthen revalidate. In order to validate a traffic simu-lation model considerable skill and persistence areneeded.

Once the model is validated, that is, it adequate-ly represents the real system, then the simulation

part of the method commences. This involves sce-nario preparation, testing the various scenarios viasimulation, evaluation of the results and implemen-tation of the emergent solution. Simulation testsshould be viewed as performing rigorous statisticalexperiments. A prerequisite for the tests is that the

simulation model is at a state where it properly rep-resents the initial state of the current traffic envi-ronment. Further, the changing input conditionswhich describe the traffic environment need to bespecified. For example, the distribution of trafficflows over the period of time where the experi-ments will be carried out need to be determined.

6 TRAFFIC MODELING AND SIMULATION OF

STROVOLOS AVENUE

As shown in Figure 3, the first step of the proposed approach is to identify and define the prob-lem. In our case the symptoms of the problemwhich are attributed to traffic congestion manifestthemselves as increasing travel times. Even thoughthe Public Works Department builds more andmore roads the traffic congestion problem becomesworse. And the reason is that we use more andmore cars for our everyday transportation as ex-

plained earlier.

The main causes for the problem of traffic con-gestion in Nicosia consist of the increasing num-

ber of vehicles and the decreasing use of the bustransportation system. Therefore the long term so-


8/12

lution to the problem is to turn around the situa-tion, that is, to decrease the number of vehicles andincrease the public transportation occupancy.

The question then becomes how do we changeour bus transportation system and make it more at-

tractive. This is what we aim to investigate in theTrafbus project concentrating on providing a faster and better quality level of service for our bus pas-sengers. The objective therefore in our modelingand simulation method is to examine various sce-narios such as dedicated bus lanes and Bus RapidTransit Systems that would provide a better levelof service for the bus transportation system. Mean-while, we need to anticipate and assess any sideeffects of to the rest of the transportation system.

Based on the stated model objectives, the devel-opment of a simulation model of Strovolos Avenue,

which is to be used as a test workbench, is carriedout. Strovolos Avenue consists of many traffic pa-rameters that need to be taken into account. Theseinclude traffic control signals, priority rules, rout-ing decisions, pedestrian crossings, signalized andunsignalised intersections and so on. A helicopter view of Strovolos Avenue simulation model is de-

picted in Figure 4 below. Strovolos Avenue is morethan 3 kilometers long extending from the borderswith Nicosia Municipality near the PresidentialPalace to the borders with Lakatatmeia Municipal-ity. Figure 4 also shows the signal times for the

various signal groups at five main signalized inter-sections of Strovolos Avenue. The depicted signal

phases correspond to the traffic peak of the morn-ing hours.

The various traffic data that we need to incorporate in our model may be classified in terms of static data and dynamic data. Static data represents

the roadway infrastructure. It includes links, whichare directional roadway segments with a specifiednumber of lanes, with start and end points as wellas optional intermediate points. Further, static dataincludes connectors between links, which are usedto model turnings, lane drops and lane gains, loca-tions and length of transit stops, position of signalheads/stop lines including a reference to the asso-ciated signal group, and positions and length of de-tectors.

Dynamic data is to be specified for traffic simu-lation applications. It includes traffic volumes in-

cluding vehicle mix (e.g. truck, HOV percentage)for all links entering the network, locations of routedecision points with routes, that is the link se-quences to be followed, differentiated by time andvehicle classification, priority rules, right-of-wayto model unsignalized intersections, permissiveturns at signalized junctions and yellow boxes or keep-clear-areas, locations of stop signs, publictransport routing, departure times and dwell times.

Having introduced the above static and dynamicdata in our model, we enter in the iterative process,which consists of model development calibration

and validation of the model. Going through sever-al iterations in developing the model, we are in a

46 AUTOMATIKA 50(2009) 1—2, 39—50


Fig. 4 The Simulation Model of Strovolos Avenue Traffic Network


9/12

AUTOMATIKA 50(2009) 1—2, 39—50 47


position to present some optimistic results concern-ing the validity of our model.

Figure 5 shows the real Vs simulated trafficflows of the various vehicle movement directionsof a central intersection of our traffic network, thatof Athalassas-Strovolou (see also figure 4). As seenin the graph below the traffic flows of real meas-

urements obtained and those of simulated results,are quite comparable. In certain intersections suchas the one shown in Figure 4, the error ranges fromonly 1% to 3%. Further, our simulation model de-monstrates the queues that we encounter in realityduring the morning peak.

Further, our simulation model demonstrates thequeues that we encounter in reality and especiallythe one at Athalassas Ave intersection where thereis no exclusive right turn lane. This is the mainqueue that a driver will encounter while goingnorth towards the centre of Nicosia between 7:00and 8:00 o’clock in the morning.

Having completed the iterative process of modeldevelopment, calibration and validation, nextcomes the preparation of scenarios, testing andevaluation of the results. Figure 6 shows a screen-shot of the simulated model as a bus has difficultyin exiting the bus bay to enter the main road.

Fig. 5 Model Validation

Fig. 6 A Snapshot of the Traffic Simulation Model


10/12

In summary the various scenarios to be evaluat-ed utilizing the developed traffic simulation modelinclude the following:

• Bus Lane with Restricted Use where dedicated

bus lanes may be assigned during the peak traf-fic hours. In this scenario based on a review of Similar systems that have been implementedworldwide an examination of the use of HighOccupancy Vehicle (HOV) lanes and issues suchas traffic flow and safety of drivers and pedestri-ans will be investigated.

• Signal pre-emption. In the Signal Pre-emptionSystem the bus is sensed, which is approachingthe traffic signal and priority is given to the bus

by extending the green phase. The system con-sists of detectors on the road that communicate

with the traffic lights. Here Global Position Sys-tems (GPS) may also be utilised.

• Extra Traffic Light. An extra traffic light with aspecial signal for buses may be installed 40—50meters from the signalised intersection in order to hold the rest of the traffic behind and give pri-ority to the bus. An extra traffic light may also

be used for the easy access of the buses from thededicated lanes to the main traffic stream. Theextra traffic lights should be synchronized withthe intersection traffic lights, which requires an-alysis of vehicle dynamics. Here bus dynamicswill be analysed concerning acceleration and de-celeration projections. Based on the bus dynam-ics analysis optimisation of the traffic flow will

be carried out.

Finally, Safety analysis based on the above sce-narios could be carried out, and the traffic flow ef-fect on the traffic network could be assessed.

Initial results show that the attractive scenarioof bus dedicated lanes could make things worse for all transport modes if not properly designed. Byexamining all possible options, and going through

detailed traffic analysis via computer simulation experimentation a viable solution is derived wherethe measures of effectiveness of travel time, aver-age speed and time delay show significant im-

provements for the bus transport mode while theimpact to rest of traffic is kept to a minimum. In

particular, the suggested solution involves a combination of central dedicated bus lanes with extratraffic lights. In fact simulation results for the bustransport mode show a reduction of total traveltime by 27%, increase average speed by 45% andreduce delay time by 28% [44]. All these improve-

ments have almost no negative effects on the other modes of transport.

7 CONCLUSIONS

Computer simulation proves to be a very power-ful tool for analyzing complex dynamical problemssuch as traffic congestions. This paper provides anoverview of the current state of research regardingtraffic simulation. Further, an approach to model-ing and simulating traffic networks is proposed andimplemented.

The proposed approach goes through variousstages, which include problem identification, modelobjectives, model development, model calibration,model validation, scenario preparation, simulationexperiments and simulated results evaluation. The

proposed approach is applied in the case of devel-oping a microscopic traffic simulation model for

the Strovolos Avenue (Nicosia, Cyprus), traffic net-work. The validation process shows that the modelsimulates traffic flows in various vehicle movementdirections intersections with an average of 95% ac-curacy.

Based on the validated microscopic simulationmodel alternative bus priority strategy scenarios in-volving dedicated bus lanes, signal preemption and

bus advance areas are evaluated. Via computer sim-ulation experimentation it was possible to developa scenario solution, which reduces total travel time

by 27 %, increase average speed by 45 % and re-

duce delay time by 28 % for the bus transport modewith almost no negative effects the rest of the traf-fic. Such simulation results allow transportationmanagers to measure the impact of their plans andtest any viable solutions prior to implementation,which represents a significant step towards design-ing optimized traffic networks.

8 ACKNOWLEDGEMENTS

The authors would like to express their appreci-

ation to the Research Promotion Foundation for funding the research project: »Modelling andAnalysis of Traffic and Impact of Dedicated BusLanes on Strovolos Avenue« AEIΦO/1104/19.Thanks also to the staff of the TransportationPlanning Section of the Public Works Departmentof the Ministry of Communication and Works of the Republic of Cyprus, for their excellent cooper-ation.

9 REFERENCES

[1] A. May, Traffic Flow Fundamentals, Prentice Hall,1990.

48 AUTOMATIKA 50(2009) 1—2, 39—50



11/12

[2] B. D. Greenshields, A Study in Highway Capacity,Highway Research Board, Proceedings, Vol. 14, 1935,

p. 458.

[3] W. F. Adams, Road Traffic Considered as a RandomSeries, J. Inst. Civil Engineers, 4, 1936, pp. 121—130.

[4] A. Bose, and P. Ioannou, Shock Waves in MixedTraffic Flow, 9th IFAC Symposium on Control inTransportation Systems, 2000.

[5] N. H. Gartner, C. Messer, A. K. Rathi, Introductionin Traffic Flow Theory, Washington, D.C.: US Fed-eral Highway Administration, 1996, pp. 1—1 to 1—4.

[6] F. L. Hall, Traffic stream characteristics in TrafficFlow Theory, Washington, D.C.: US Federal HighwayAdministration, 1996, pp. 2—1 to 2—34.

[7] R. J. Koppa, Human Factors in Traffic FlowTheory, Washington, D.C.: US Federal Highway Ad-ministration, 1996, pp. 3—1 to 3—31.

[8] R. W. Rothery, Car Following Models in TrafficFlow Theory, Washington, D.C.: US Federal HighwayAdministration, 1996, pp. 4—1 to 4—42.

[9] R. Kuhne, and P. Michalopoulos, Continuum FlowModels in Traffic Flow Theory, Washington, D.C.:US Federal Highway Administration, 1996, pp. 5—1to 5—51

[10] J. C. Williams, Macroscopic Flow Models in TrafficFlow Theory, Washington, D.C.: US Federal HighwayAdministration, 1996, pp. 6—1 to 3—31.

[11] S. Ardekani, E. Hauerand and E. Jamei, TrafficImpact Models in Traffic Flow Theory, Washing-

ton, D.C.: US Federal Highway Administration, 1996, pp. 7—1 to 7—24.

[12] R. J. Troutbeck, and W. Brilon, Unsignalized Inter-section Theory in Traffic Flow Theory, Washington,D.C.: US Federal Highway Administration, 1996, pp.8—1 to 8—47.

[13] N. Rouphail, A. Tarko and J. Li, Traffic Flow atSignalized Intersections in Traffic Flow Theory,Washington, D. C.: US Federal Highway Administra-tion, 1996, pp. 9—1 to 9—32.

[14] E. Lieberman, and A. J. Rathi, Traffic Simulation inTraffic Flow Theory, Washington, D.C.: US FederalHighway Administration, 1996, pp. 10—1 to 3—23.

[15] K. Nagel, Partical hopping models and traffic flowtheory. Physics Review E., (53):4655—4672, 1996.

[16] K. Nagel & M. Schreckenberg, A cellular automa-ton model for freeway traffic. J. Phys. I France 2,1992 (12): 2221—2229.

[17] T. Waldeer, Kinetic Theory in Vehicular TrafficFlow Modeling, Proc. of 25. Conference in RarefiedGas Dynamics (RGD), St. Petersburg, Russia (2006).

[18] T. Waldeer, Numerical Investigation of a MesoscopicVehicular Traffic Flow Model Based on a Stochas-tic Acceleration Process, Transp. Theory Stat. Phys.,2004.

[19] PTV AG, VISSIM Version 4.10, User Manual, March2005.

[20] P. Byungkyu, W. Jongsun, Y. Ilsoo, Application of Microscopic Simulation Model Calibration and Val-idation Procedure: Case Study of CoordinatedActuated Signal System, TRR, 2006.

[21] G. Gomes, A. May, R. Horowitz, Congested Freeway

Microsimulation Model using Vissim, TRR, 2004.

[22] M. Fellendorf, and P. Vortisch, Validation of theMicroscopic Traffic Flow Model VISSIM in Differ-ent Real-World Situations, Paper presented at theAnnual Meeting, TRB, Washington, DC, 2001.

[23] FHWA, CORSIM 5.1, User's Guide and ReferenceManual, February 2003.

[24] S. Lin, 1998. CORSIM micro-node logic, TechnicalReport, Federal Highway Administration, McLean,VA.

[25] L. Zhang, G. McHale, Y, Zhang, Modeling andValidating Corsim Freeway Origin-Destination

Volumes, TRR, 2003[26] P. D. Prevedouros, Y. Wang, Simulation of large free-

way and arterial network with Corsim, integration,and Watsim, TRR, 1999.

[27] TSS, AIMSUN-NG Version 5.0, User's Manual, November 2005.

[28] J. Barcelo, et al., AIMSUN modeling issues, Trans- port Simulation Systems Internal Report (www.tss-bcn.com) 1998.

[29] J. Barcelo, J. Casas, Stochastic Heuristic DynamicAssignment Based on AIMSUN Microscopic TrafficSimulator, TRR, 11th IFAC Symposium on Control

in Transportation Systems 2006.[30] J. Barcelo, A hybrid simulation framework for ad-

vanced transportation analysis, 11th IFAC Sympo-sium on Control in Transportation Systems 2006.

[31] Quadstone Limited, PARAMICS, Version 5.1, Model-er User Guide, 2005).

[32] K. Ozbay, J. Bartin, O. Bekir; S. Mudigonda, Micro-scopic Simulation and Calibration of IntegratedFreeway and Toll Plaza Model, TRR, 2006.

[33] Y. Gardes, Evaluating Traffic Calming and Cap-acity Improvements on SR-20 Corridor UsingMicroscopic Simulation, TRR, 2006.

[34] C. Jacob, B. Abdulhai, Automated Adaptive TrafficCorridor Control Using Reinforcement Learning:Approach and Case Studies, TRR 2006.

[35] G. Papageorgiou, Towards a microscopic simulationmodel for traffic management: a computer basedapproach, 11th IFAC Symposium on Control in Trans-

portation Systems, 2006.

[36] G. Papageorgiou, P. Damianou, A. Pitsillides, T.Aphames, P. Ioannou, A Microscopic Traffic Simula-tion Model for Transportation Planning in Cyprus,International Conference on Intelligent Systems AndComputing: Theory And Applications (ISYC), July6—7, 2006, Cyprus.

[37] The SMARTEST project, http://www.its.leeds.ac.uk/ projects/smartest/

AUTOMATIKA 50(2009) 1—2, 39—50 49



12/12

[38] Bloomberg and Dale, Comparison of vissim and cor-sim traffic simulation models on a congested net-work , TRR, 2000.

[39] S. A. Boxill and L. Yu, An Evaluation of TrafficSimulation Models for Supporting ITS Develop-

ment, Technical Report 167602-1. Texas SouthernUniversity, 2000.

[40] Sakda Panwai and Hussein Dia, ComparativeEvaluation of Microscopic Car-Follo BehaviorIEEE Transactions on Intelligent TransportationSystems, vol. 6, no. 3, September 2005.

[41] D. Manstetten, W. Krautter and T. Schwab, Trafficsimulation supporting urban control system devel-

opment, in 4th World Congr. Intelligent TransportSystem, Berlin, Germany, 1997, pp. 1—8.

[42] Fred Choa, Ronald T. Milam, David Stanek, COR-SIM, PARAMICS, and VISSIM: What the Man-uals Never Told You, 9th TRB conference on the ap-

plication of transportation planning methods, 2003.

[43] Highway Capacity Manual, FHWA, 2000.

[44] G. Papageorgiou, P. Damianou, A. Pitsillides, T.Aphames, D. Charalambous, P. Ioannou, TRAFBUS

— Modelling and Analysis of Traffic and Impact of Dedicated Bus Lanes on Strovolos Avenue, IpsipetisJournal of the Cyprus Research Promotion Founda-tion, Issue 17, 2008.

50 AUTOMATIKA 50(2009) 1—2, 39—50


Modeliranje i simulacija transportnih sustava: pristup planiranja scenarija. Postaje sve razvidnije dase izgradnjom novih cesta ne moè rije{iti problem prometne zagu{enosti, ve} se ~ak moè i pove}ati. Dakle,

umjesto pro{irenja mreè cesta trebalo bi uloìti energiju u pove}anje razine usluga javnog transporta, poglavi-to primjenom novih tehnologija kao {to su ra~unalna i informacijska tehnologija. Napredne tehnologije uklju~e-ne u tzv. Inteligentne transportne sustave (ITS) pruàju velike mogu}nosti za ublaàvanje problema prometnezagu{enosti. S druge strane, ITS tehnologije zahtijevaju rigorozne provjere i vrednovanja koja se jedino mogu provesti pomo}u ra~unalnih simulacija. U radu se daje pregled na~ina simulacije prometa te proces razvojamikroskopskog simulacijskog modela jako zagu{ene prometne mreè u Nikoziji na Cipru. Vrednovani simula-cijski model pruà planerima transporta i upraviteljima prometa mogu}nost provjeravanja raznih rje{enja sce-narija brzog autobusnog prometa, uklju~uju}i primjenu inteligentnih prometnih sustava prije njihove imple-mentacije.

Klju~ne rije~i: modeliranje, simulacija, transportni sustavi, brzi autobusni promet

AUTHORS’ ADDRESSES:

George Papageorgiou1, Pantelis Damianou1,Andreas Pitsillides1, Thrasos Aphamis2,Demetris Charalambous3, Petros Ioannou4

1 University of Cyprus, Department of Mathematics,Cypruse-mail: [email protected] (George Papageorgiou)

2 Transportation Planning Section, Public Works Depart-ment, Ministry of Communication and Works, Cyprus

3 University of Lancaster, Department of Physics, UK 4 University of Southern California, Center for Advanced

Transportation Technologies, CA, USAReceived: 2008-07-14Accepted: 2008-11-17

modelling_and_simulation_of_transportation_systems_a_scenario_planning_approach-libre (1).pdf

Documents