Advanced Traffic Simulation Framework for Networked Driving Simulators

Preprints, 8th IFAC International Symposium on Advances in Automotive Control Advances in Automotive Control Symposium Preprints, 8th IFAC International Preprints, 8th IFACNorrköping, International Symposium on on June 19-23, 2016. Sweden Preprints, 8th IFACNorrköping, International Symposium ononline at www.sciencedirect.com June 19-23, Sweden Advances in Automotive Control Available Advances in2016. Automotive Control Advances in2016. Automotive Control June 19-23, Norrköping, Sweden June 19-23, 2016. Norrköping, Sweden June 19-23, 2016. Norrköping, Sweden

ScienceDirect

IFAC-PapersOnLine 49-11 (2016) 101–108 Advanced Framework for Advanced Traffic Traffic Simulation Simulation Framework for Networked Networked Driving Driving Simulators Simulators Advanced Traffic Simulation Framework for Networked Driving Simulators Advanced Simulation Framework for Driving Simulators Advanced Traffic Traffic Simulation Framework forS.Networked Networked K. Abdelgawad, S. Henning, P. Biemelt, Gausemeier, A. Driving Trächtler Simulators

K. Abdelgawad, S. Henning, P. Biemelt, S. Gausemeier, A. Trächtler  K. S.  K. Abdelgawad, Abdelgawad, S. S. Henning, Henning, P. P. Biemelt, Biemelt, S. Gausemeier, Gausemeier, A. A. Trächtler Trächtler K. Abdelgawad, S. Henning, P. Biemelt, S. Gausemeier, A. Trächtler  University of Paderborn, Paderborn, 33102, Germany  University of Paderborn, Paderborn, 33102, Germany Tel: {Kareem.Abdelgawad, Sven.Henning, Patrick.Biemelt, Sandra.Gausemeier, University of Paderborn, Paderborn, 33102, Germany Tel: +49(0)5251/606228; +49(0)5251/606228; e-mail: e-mail: {Kareem.Abdelgawad, Sven.Henning, Patrick.Biemelt, Sandra.Gausemeier, University of Paderborn, Paderborn, 33102, Germany Ansgar.Trächtler}@hni.upb.de. University of Paderborn, Paderborn, 33102, Germany Tel: +49(0)5251/606228; e-mail: {Kareem.Abdelgawad, Sven.Henning, Patrick.Biemelt, Sandra.Gausemeier, Ansgar.Trächtler}@hni.upb.de. Tel: +49(0)5251/606228; e-mail: {Kareem.Abdelgawad, Sven.Henning, Patrick.Biemelt, Tel: +49(0)5251/606228; e-mail: {Kareem.Abdelgawad, Sven.Henning, Patrick.Biemelt, Sandra.Gausemeier, Sandra.Gausemeier, Ansgar.Trächtler}@hni.upb.de. Ansgar.Trächtler}@hni.upb.de. Ansgar.Trächtler}@hni.upb.de. Abstract: Abstract: Simulating Simulating surrounding surrounding vehicles vehicles in in aa driving driving simulator simulator is is essential essential for for drivers drivers to to experience experience realistic traffic situations. However, simulating traffic several miles ahead or behind a driving simulator Abstract: Simulating surrounding vehicles in a driving simulator is essential for drivers to experience realistic traffic situations. However, simulating traffic several miles ahead or behind a driving simulator Abstract: Simulating surrounding vehicles in aaconcept driving simulator is essential for drivers to experience vehicle is not efficient. This paper presents the and structure of a traffic simulation framework Abstract: Simulating surrounding vehicles in driving simulator is essential for drivers to experience realistic traffic situations. However, simulating traffic several miles ahead or behind a driving simulator vehicle is not efficient. This paper presents the concept and structure of a traffic simulation framework realistic traffic situations. However, simulating traffic several miles or behind aa to driving simulator for of driving simulators. The framework enables users select target realistic traffic situations. However, simulating traffic several miles ahead ahead behind driving simulator vehicle is not efficient. This paper presents the concept and structure of aa or traffic simulation framework for an an environment environment of networked networked driving simulators. The framework enables users to select aa target vehicle is not efficient. This paper presents the concept and structure of traffic simulation framework simulator vehicle and generate traffic only in its neighborhood. The other participating drivers are vehicle is not efficient. This paper presents the concept and structure of a traffic simulation framework for an environment of networked driving simulators. The framework enables users to select a target simulator vehicle and generate traffic onlysimulators. in its neighborhood. The enables other participating drivers are for an environment of networked driving The framework users to select aa target considered, so that they experience realistic traffic behaviour as well. The developed framework has been for an environment of networked driving simulators. The framework enables users to select target simulator vehicle and generate traffic only in its neighborhood. The other participating drivers are considered,vehicle so that and they generate experience realistic traffic behaviour as well.The The other developed framework has been simulator traffic only in neighborhood. participating drivers are validated aathat demonstrator to observe the interaction of the generated traffic. simulator vehicle and generate traffic onlytraffic in its itsbehaviour neighborhood. participating drivers are considered, so they experience realistic as well. The developed framework has been validated with with demonstrator to graphically graphically observe the interaction ofThe the other generated traffic. considered, so that they experience realistic traffic behaviour as well. The developed framework has been considered, so that they experience realistic traffic behaviour as well. The developed framework has been validated with aa demonstrator to observe the interaction generated traffic. © 2016, IFAC (International Federation of Automatic Control) Hosting of bythe Elsevier Ltd. All rights reserved. validated with to graphically graphically observe the of generated traffic. Keywords: Vehicle simulators, models, Road Distributed simulation, validated a demonstrator demonstrator graphically observe the interaction interaction of the the generatedTraining traffic. Keywords:with Vehicle simulators,toDriver Driver models, Road traffic, traffic, Distributed simulation, Training Keywords: Vehicle simulators, simulators, Driver models, models, Road traffic, Distributed Distributed simulation, simulation, Training Training Keywords:  traffic, Keywords: Vehicle Vehicle simulators, Driver Driver models, Road Road  traffic, Distributed simulation, Training repetitive repetitive and and predefined predefined scenarios. scenarios. They They lack lack realism realism and and 1.  1. INTRODUCTION INTRODUCTION multi-interactivity required for future automotive  repetitive and predefined scenarios. They lack realism and multi-interactivity required for future automotive repetitive and predefined scenarios. They lack realism and 1. INTRODUCTION applications. repetitive and predefined scenarios. They lack realism and 1. INTRODUCTION Driving is one of the most popular daily activities that people multi-interactivity required for future automotive applications. INTRODUCTION required for future automotive Driving is one of the1.most popular daily activities that people multi-interactivity multi-interactivity required for future automotive perform. Nevertheless, it is complex and relatively applications. Driving is one of the most activities people applications. perform. Nevertheless, itpopular is aa daily complex and that relatively In Driving is one of the most popular daily activities that people applications. In networked networked driving driving simulation, simulation, different different human-driven human-driven dangerous activity. Drivers have to interact with different Driving is one of the most popular daily activities that people perform. Nevertheless, Nevertheless, it is is complex and relatively dangerous activity. Drivers haveaa to interact and with relatively different vehicles in a distributed simulation environment can perform. it complex In networked driving simulation, different human-driven vehicles in a distributed simulation environment can vehicle like, e.g., transportation systems, perform. Nevertheless, is a to complex and relatively networked driving simulation, different human-driven dangerous activity. have interact with different vehicle systems, systems, like,Drivers e.g.,it intelligent intelligent transportation systems, In participate and interact within a common traffic scenario In networked driving simulation, different human-driven dangerous activity. Drivers have to interact with different vehicles in a distributed simulation environment can participate and interact within a common traffic scenario advanced driver assistance systems (ADAS), and in-vehicle dangerous activity. Drivers have to(ADAS), interact and within-vehicle different vehicles in aa distributed simulation environment can vehicle systems, like, e.g., intelligent transportation systems, advanced driver assistance systems (Oeltze and Schießl (2015)). The ability to create aascenario virtual vehicles inand distributed simulation environment can vehicle systems, like, e.g., intelligent transportation systems, participate interact within a common traffic (Oeltze and Schießl (2015)). The ability to create virtual information systems (Pauzie (1994)). These systems vehicle systems, like, e.g., intelligent transportation systems, participate and interact within a common traffic scenario advanced driver assistance systems (ADAS), and in-vehicle information systems (Pauzie (1994)). These systems participate driving environment simultaneously accessed by two or more and interact within a common traffic scenario advanced driver assistance systems (ADAS), and in-vehicle (Oeltze environment and Schießl Schießl simultaneously (2015)). The The ability ability to create create virtual driving accessed by twoaaorvirtual more influence drivers’ behaviour and ability to handle traffic advanced driver assistance systems (ADAS), in-vehicle and (2015)). to information systems (Pauzie (1994)). systems influence drivers’ behaviour and their their ability These toand handle traffic (Oeltze human drivers allows aa much closer approximation of (Oeltze and Schießl (2015)). The ability to create aorreality, virtual information systems (Pauzie (1994)). These systems driving environment simultaneously accessed by two more human drivers allows much closer approximation of reality, situations. information systems (Pauzie (1994)). These systems driving environment simultaneously accessed by two or more influence drivers’ behaviour and their ability to handle traffic situations. with its attendant risks and uncertainty. Networked driving driving environment simultaneously accessed by two or more influence drivers’ behaviour and their ability to handle traffic human drivers allows a much closer approximation of reality, with its attendant risks and uncertainty. Networked driving influence drivers’ behaviour and their ability to handle traffic human drivers allows a much closer approximation of reality, situations. simulation can be used for future research and development human drivers allows a much closer approximation of reality, situations. Automotive manufacturers and suppliers focus on with its attendant risks and uncertainty. Networked driving simulation can be used for future researchNetworked and development situations. its attendant risks and uncertainty. driving Automotive manufacturers and suppliers focus on with areas, serve applications involving with its which attendant risks automotive andfuture uncertainty. Networked driving overcoming new technological challenges and addressing simulation can be used for research and development areas, which serve automotive applications involving Automotive manufacturers and suppliers focus on simulation can be used for future research and development overcoming new technological challenges and addressing Automotive manufacturers and suppliers focus on connectivity and interaction. Examples for these applications simulation can be used for future research and development acceptance for vehicles (Lu et al. (2014)) Automotive manufacturers and suppliers on connectivity areas, which automotive applications involving andserve interaction. Examples for these applications overcomingaspects new technological challenges and addressing areas, which automotive applications involving acceptance aspects for connected connected vehicles (Lu et focus al. (2014)) overcoming new technological challenges and addressing are: development of connected vehicle systems (Tideman and areas, whichandserve serve automotive applications involving and future ADAS. From a technology-driven perspective, overcoming new technological challenges and addressing connectivity interaction. Examples for these applications are: development of connected vehicle systems (Tideman and acceptance aspects for connected vehicles (Lu et al. (2014)) connectivity and interaction. Examples for these applications and future aspects ADAS. for From a technology-driven perspective, acceptance connected vehicles (Lu et al. (2014)) van Noort (2013)), conjoint training of drivers (Kandhai et al. connectivity and interaction. Examples for these applications these systems must be evaluated regarding their functionality, acceptance aspects for connected vehicles (Lu et al. (2014)) are: development of connected vehicle systems (Tideman and van Noort (2013)), conjoint training of drivers (Kandhai et al. and future ADAS. From a technology-driven perspective, are: development of connected vehicle systems (Tideman and thesefuture systems must beFrom evaluated regarding their functionality, and ADAS. aa technology-driven perspective, (2011)), and investigations of drivers’ cooperative behaviour are: development of connected vehicle systems (Tideman and robustness, and interoperability with systems provided by and future ADAS. From technology-driven perspective, van Noort (2013)), conjoint training of drivers (Kandhai et al. (2011)), and investigations of drivers’ cooperative behaviour these systems must be evaluated regarding their functionality, van Noort (2013)), conjoint training of drivers (Kandhai et robustness, and interoperability with systems provided by van these systems must be regarding their functionality, (Mayenobe et al. (2004)). Noortand (2013)), conjoint of training of cooperative drivers (Kandhai et al. al. different manufacturers. From the human factor point of these systems must be evaluated evaluated regarding their functionality, (2011)), investigations drivers’ behaviour (Mayenobe et al. (2004)). robustness, and interoperability with systems provided by (2011)), and investigations of drivers’ cooperative behaviour different manufacturers. From the human factor point of robustness, and interoperability with systems provided by (2011)), andetinvestigations of drivers’ cooperative behaviour view, it is important to study the impact of these systems on robustness, and interoperability with systems provided by (Mayenobe al. (2004)). different From human factor point of (Mayenobe et al. view, it ismanufacturers. important to study thethe impact of these systems on Simulating surrounding different manufacturers. From the human factor point of al. (2004)). (2004)). vehicles Simulating et surrounding vehicles in in aa driving driving simulator simulator is is drivers and their cooperative behaviour (Mayenobe et al. different manufacturers. From the human factorsystems point of (Mayenobe view, it is important to study the impact of these on drivers and their cooperative behaviour (Mayenobe et al. essential to let simulator drivers experience realistic traffic view, it is important to study the impact of these systems on Simulating surrounding vehicles in aa driving simulator is essential to let simulator drivers experience realistic traffic (2004)). Further, drivers must learn how to use these systems view, it is important to study the impact of these systems on Simulating surrounding vehicles in driving simulator is drivers and their cooperative behaviour et al. (2004)). Further, drivers must learn how to (Mayenobe use these systems situations (Punzo and Ciuffo (2010)). characteristics of Simulating surrounding vehicles in a The driving simulator is drivers and their cooperative behaviour (Mayenobe et al. essential to let simulator drivers experience realistic traffic situations (Punzo and Ciuffo (2010)). The characteristics of and not to overestimate their capabilities in order to achieve drivers and their cooperative behaviour (Mayenobe et al. essential to let simulator drivers experience realistic traffic (2004)). Further, drivers their must capabilities learn how how to toinuse use these systems and not to overestimate order to systems achieve essential surrounding traffic influence the driver’s behaviour, and to let simulator drivers experience realistic traffic (2004)). Further, drivers must learn these situations (Punzo Ciuffo (2010)). The characteristics of surrounding trafficand influence the driver’s behaviour, and safe traffic flow. (2004)). Further, must capabilities learn how toinuse these systems situations (Punzo and Ciuffo (2010)). The characteristics of and to overestimate order to achieve safe not traffic flow. drivers their hence, the ability to drive the vehicle. However, simulating situations (Punzo and Ciuffo (2010)). The characteristics of and not to overestimate their capabilities in order to achieve surrounding traffic influence the driver’s behaviour, and hence, the ability to drive the vehicle. However, simulating and not to overestimate their capabilities in order to achieve surrounding traffic influence the driver’s behaviour, and safe traffic flow. traffic several miles ahead or behind the driving simulator surrounding traffic influence the driver’s behaviour, and safe traffic flow. In this regard, driving simulation and traffic simulation are hence, the ability to drive However, several miles aheadthe or vehicle. behind the drivingsimulating simulator safe traffic flow.driving simulation and traffic simulation are traffic hence, the ability to the However, simulating In this regard, vehicle is not efficient from aa vehicle. computational point of hence, the ability to drive drive the vehicle. However, simulating used in the automotive field. In driving simulation, a traffic several miles ahead or behind the driving simulator vehicle is not efficient from computational point of view. view. In this regard, driving simulation and traffic simulation are traffic several miles ahead or behind the driving simulator used inregard, the automotive field. In driving simulation,area traffic In this driving simulation and traffic simulation Previous approaches have been developed to efficiently several miles ahead or behind the driving simulator participant controls a virtual vehicle in a simulated In this regard, driving simulation and traffic simulation are vehicle is not efficient from a computational point of view. Previous approaches have been developed to efficiently used in the automotive field. In driving simulation, a vehicle is not efficient from a computational point of view. participant controls a virtual vehicle in a simulated used in the automotive field. In driving simulation, aa vehicle generate traffic vehicles for a single driving simulator is not efficient from a computational point of view. environment. In traffic simulation, driver models are used to used in the automotive field. In driving simulation, Previous approaches have been developed to efficiently generate traffic vehicles for a single driving simulator participant controls a virtual vehicle in a simulated Previous approaches have been developed to efficiently environment. In traffic simulation, driver models are used to participant controls aa virtual vehicle in aa simulated (Olstam and Lundgren (2008)). This paper presents the Previous approaches have been developed to efficiently investigate aspects related to a whole traffic system. Driving participant controls virtual vehicle in simulated generate traffic vehicles for a single driving simulator (Olstam and Lundgren (2008)). This paper presents the environment. In traffic simulation, driver models are used to generate traffic vehicles for aa framework single driving simulator investigate aspects related to a whole traffic system. Driving environment. In traffic simulation, driver models are used to concept of a traffic simulation for networked generate traffic vehicles for single driving simulator simulation and simulation are used together create environment. Intraffic traffic simulation, driver models areto used to concept (Olstam and Lundgren (2008)). This paper presents the of a traffic simulation framework for networked investigate aspects related to a whole traffic system. Driving (Olstam and Lundgren (2008)). This paper presents the simulation and traffic simulation are used together to create investigate aspects related to aa whole traffic system. Driving driving simulators. The developed framework allows users, (Olstam anda Lundgren (2008)). framework This paperfor presents the scenarios that involve human-driven and programmed investigate aspects related to whole traffic system. Driving concept of traffic simulation networked driving simulators. The developed framework allows users, simulation and and traffic simulation are used used and together to create create concept of a traffic simulation framework for networked scenarios that traffic involve human-driven programmed simulation simulation are together to i.e., operators or developers, to select a target simulator concept of a traffic simulation framework for networked vehicles and Ciuffo (2010)). this provides simulation and traffic simulation are However, used and together to create i.e., driving simulators. The developed framework allows users, operators or developers, to select a target simulator scenarios that involve human-driven programmed driving simulators. The developed framework allows users, vehicles (Punzo (Punzo and Ciuffo (2010)). However, this provides scenarios that involve human-driven and programmed vehicle and generate traffic with realistic interactions only in driving simulators. The developed framework allows users, only an approximation of the level of uncertainty and scenarios that involve human-driven and programmed i.e., operators or developers, to select a target simulator vehicle and generate traffic with realistic interactions only in vehicles (Punzo and Ciuffo (2010)). However, this provides operators or developers, to select a target simulator only an (Punzo approximation of (2010)). the level of uncertainty and i.e., vehicles and Ciuffo However, this provides its neighbourhood. In other words, traffic vehicles are i.e., operators or developers, to select a target simulator unpredictability encountered when multiple human drivers vehicles (Punzo and Ciuffo (2010)). However, this provides vehicle and generate traffic with realistic interactions only in its neighbourhood. In other words, traffic vehicles are only an approximation of the level of uncertainty and vehicle and generate traffic with realistic interactions only in unpredictability encountered whenlevel multiple human drivers only an of the of uncertainty and generated within sight distance or area of interest of the vehicle andonly generate traffic withwords, realistic interactions only in are operating in the same as the case in only an approximation approximation ofenvironment, the level ofi.e., uncertainty and its neighbourhood. In other traffic vehicles are generated only within sight distance or area of interest of the unpredictability encountered when multiple human drivers its neighbourhood. In other words, traffic vehicles are are operating in the same environment, i.e., as the case in unpredictability encountered when multiple human drivers target driving simulator vehicle. The other participating its neighbourhood. In other words, traffic vehicles are real-world traffic situations (Muehlbacher (2015)). In other unpredictability encountered when multiple human drivers generated only within sight distance or area of interest of the target driving simulator vehicle. The other participating are operating in the same environment, as the Incase in generated only sight distance or area of of real-world traffic situations (Muehlbacheri.e., (2015)). other are operating in same i.e., as case in human drivers are considered, they experience only within within sightvehicle. distanceso orthat area of interest interest of the the words, individual simulators serve only as isolated are operating in the the same environment, environment, i.e., as the thesolutions, case in generated target driving simulator The other participating human drivers are considered, so that they experience real-world traffic situations (Muehlbacher (2015)). In other target driving simulator vehicle. The other participating words, individual simulators serve only as isolated solutions, real-world traffic situations (Muehlbacher (2015)). In other realistic traffic behaviour as well. The main objective is target driving simulator vehicle. The other participating which enable research, development, and drivers training in real-world traffic situations (Muehlbacher (2015)). In other human drivers are considered, experience traffic behaviour as well. so Thethat mainthey objective is to to words, individual simulators serve only as isolated solutions, human drivers are considered, so that they experience which enable research, development, and drivers training in realistic words, individual simulators serve only as isolated solutions, simulate traffic vehicles and to let all simulator drivers human drivers are considered, so that they experience words, individual simulators serve only as isolated solutions, realistic traffic behaviour as well. The main objective is simulate traffic vehicles and to let all simulator drivers which enable enable research, development, development, and drivers drivers training in in realistic traffic behaviour as well. The main objective is to which traffic behaviour as well. Theall main objective is to to which enable research, research, development, and and drivers training training in realistic simulate traffic vehicles and to let simulator drivers simulate traffic vehicles and to let all simulator drivers Copyright © 2016 IFAC 103 simulate traffic vehicles and to let all simulator drivers Copyright © 2016, 2016 IFAC 103 Hosting by Elsevier Ltd. All rights reserved. 2405-8963 © IFAC (International Federation of Automatic Control) Copyright ©under 2016 responsibility IFAC 103Control. Peer review© of International Federation of Automatic Copyright 2016 IFAC 103 Copyright © 2016 IFAC 103 10.1016/j.ifacol.2016.08.016

IFAC AAC 2016 102 June 19-23, 2016. Norrköping, Sweden

K. Abdelgawad et al. / IFAC-PapersOnLine 49-11 (2016) 101–108

undergo realistic behaviour at the same time. Moreover, users are able to vary the speed and density of the generated traffic during simulation runtime as desired. The developed framework has been tested with a demonstrator to observe the interaction of the generated traffic.

capacity. Therefore, some commercial solutions define a simulation area around the driving simulator vehicle, where traffic objects are generated. For example, v-TRAFFIC from VIRES Simulation Technology is a traffic and scenario simulation engine that can be used with driving simulators to simulate surrounding traffic objects (Neumann-Cosel et al. (2009)). Principally, there is no limit for the number of traffic objects that can be created. Moreover, v-TRAFFIC utilizes a simulation area/window concept to reduce computation effort.

This paper is structured as follows: Section 2 presents related work in the field of traffic simulation for driving simulators. Section 3 introduces the concept and structure of the developed traffic simulation framework. Section 4 describes the design and functionality of each subsystem. The validation of the framework is discussed in Section 5. Finally, Section 6 presents potential future work for the developed framework.

Despite promising work in the research and commercial fields, there is still no traffic simulation model that can be used efficiently when multiple driving simulators are connected in the same virtual environment. Previous work and existing solutions consider only setups consisting of one driving simulator. This paper presents a concept of a traffic simulation framework that considers a driving simulation setup, where multiple human drivers interact together in the same virtual scenario. The following section introduces the concept and the structure of the developed traffic simulation framework.

2. RELATED WORK According to literature review, there is a lot of research work in the field of traffic simulation for driving simulators. For example, a symbolic vision model for traffic simulation is presented in (Espie and Auberlet (2007)). This model allows each traffic vehicle to observe various elements of its environment, like, e.g., other traffic vehicles, road signs, lane markings, etc. Hence, each traffic object can adjust its behaviour based on the traffic situation as well as the perceived road environment. An approach using Hierarchical Concurrent State Machines (HCSM) is introduced in (Bonakdarian et al. (1998)). This model allows each traffic vehicle to check the situation, and hence, to automatically switch between different decisions, like, e.g., lane tracking or change, object following, traffic collision avoidance, static object collision avoidance, etc. A model for traffic simulation that utilizes a fuzzy logic approach is presented in (Wright et al. (2002)). By introducing a degree of indeterminism, this model mimics the behaviour of a typical human driver. To reduce computation effort, a concept for traffic vehicles simulation in a specified candidate area is introduced in (Olstam (2003)). A comprehensive survey for previous traffic simulation approaches is presented in (Olstam (2005a)).

3. CONCEPT AND STRUCTURE OF THE TRAFFIC SIMULATION FRAMEWORK A crucial difference between traffic simulation for driving simulators and other purposes of traffic simulation is that one of the vehicles is controlled by a human driver. Consequently, traffic vehicles have not only to react to each other, but also to the behaviour of the simulator driver. The developed traffic simulation framework in this work considers networked driving simulation setups. That is, the generated traffic vehicles are shared among multiple human drivers that interact within the same virtual scenario. Fig. 1 shows the structure of the developed traffic simulation framework along with the main building blocks of the overall simulation environment. Traffic simulator

SUMO from DLR Institute of Transportation Systems is an open and flexible suite for modelling traffic systems including road vehicles and pedestrians (Behrisch et al. (2011)). To simulate surrounding traffic for driving simulators, there are several commercial tools in the market. For instance, ASM Traffic from dSPACE is a framework consisting of Simulink models to simulate road traffic (Amelunxen (2015)). The model simulates one test vehicle and up to 15 independent traffic vehicles and it can be utilized for interactive driving simulations. The ASM traffic editor enables users to define different traffic scenarios through its graphical user interface. IPGTraffic from IPG Automotive provides models to represent the interactions of traffic vehicles for driving simulators (Miquet et al. (2010)). Unlimited number of traffic objects can be created, where maneuvers can be assigned to each traffic object individually. One problem with most commercial solutions is that they simulate an entire geographic area. For wide areas or long roads, many traffic vehicles have to be simulated, especially when running long driving simulator experiments. This may introduce challenges regarding the required computation

Traffic initialisation

Driver model

Dynamic windows generation

Traffic vehicles module Traffic vehicle 1

Traffic observer

Traffic vehicle 2

Vehicles generation/discarding

...

Traffic speed/density controller

Traffic vehicle N

Simulation management station

Driving simulator 1

Driving simulator 2

...

Driving simulator N

Fig. 1: The developed traffic simulation framework and its relation with the networked driving simulation environment. 104

IFAC AAC 2016 June 19-23, 2016. Norrköping, Sweden

K. Abdelgawad et al. / IFAC-PapersOnLine 49-11 (2016) 101–108

The traffic initialisation function outputs six signals according to user preferences. These will be described subsequently while discussing the corresponding subsystems.

Principally, the traffic simulation framework is designed for an environment of networked driving simulators. All participating driving simulators interact with the same generated traffic. Although one simulator vehicle is selected for traffic simulation, the other participating driving simulators are considered, so that they experience realistic traffic behaviour. Traffic vehicles are simulated only within sight distance of simulator vehicle(s) to reduce required computation capacity. Moreover, the speed and density of generated traffic can be adjusted as desired.

4.2 Dynamic windows generation Instead of simulating an entire geographical area, models and approaches have been developed to generate traffic vehicles only in the closest neighbourhood of a simulator vehicle (Olstam (2005b)). In this work, an environment of multiple driving simulators is considered. Traffic vehicles are simulated only within the neighbourhood area of a selected simulator vehicle, i.e., target simulator vehicle. They are discarded once they leave this neighbourhood area. This neighbourhood area is centred on the target simulator vehicle and it moves with the same speed; it is called ‘simulation window’. To avoid sudden appearance and disappearance of traffic vehicles, the simulation window is designed to be at least as long as the sight distance ahead and behind the target simulator vehicle.

The traffic simulation framework receives information about the position, orientation, and speed of each participating driving simulator. The positions and orientations of simulated traffic vehicles represent the main output of the framework. These outputs are received by the so-called simulation management station. This can be, for example, a computer with runtime infrastructure for High Level Architecture (Dahmann (1999)), which manages data forwarding and synchronisation between distributed simulation participants. Description for this simulation management station is beyond the scope of this work. The output of the traffic simulation framework is forwarded to the relevant driving simulator(s). The following sub-sections discuss the design of the functional units of the developed traffic simulation framework and their main input/output signals.

In general, sight distance in road design is defined as the length of roadway visible to a driver (Hang et al. (2008)). As the case in real life, road geometry in a 3D environment is one of the factors affecting the sight distance available to simulator drivers. For the sake of simplicity in this work, a road course without intersections or ramps was chosen; it is a straight segment with two lanes at each direction.

4. FUNCTIONS AND SUBSYSTEMS DESIGN As shown in Fig. 1, the developed traffic simulation framework consists of subsystems or functional units. These were developed in Matlab/Simulink and arranged in a modular structure, so that they communicate in a loosely coupled fashion. The following sub-sections describe the design and functionality of each subsystem together with its relationship with other subsystems.

The function of dynamic windows generation receives the positions of all participating simulator vehicles, a signal indicating the target simulator vehicle, and the set sight distance. The latter two signals are received from the traffic initialisation function. The function does not only define a neighbourhood area for the target simulator vehicle, but also for each participating simulator vehicle. In this case, neighbourhood areas are called ‘moving windows’. Fig. 3 shows that main input/output signals of the dynamic windows generation function.

4.1 Traffic initialisation The traffic initialisation function represents the interface between users and the traffic simulation framework. Framework users can set some parameters, like, e.g., drivers’ sight distance, traffic density level, etc., at the beginning of the simulation session as well as during simulation runtime. Fig. 2 shows the traffic initialisation function and its outputs.

User preferences

Traffic initialisation

103

Set headway distance Set sight distance Target simulator vehicle Traffic speed level Traffic density level Traffic veh. length

Fig. 3: Dynamic windows generation and its input/output signals. Due to the simplicity of the adopted road course in this work, each moving window is defined with two virtual parallel lines at the sight distance behind and in front of the simulator vehicle (see Fig. 4). The sight distance is variable, i.e., it can be adjusted as desired through the traffic initialisation function.

Fig. 2: Traffic initialisation function and its input/output signals. 105

IFAC AAC 2016 104 June 19-23, 2016. Norrköping, Sweden

K. Abdelgawad et al. / IFAC-PapersOnLine 49-11 (2016) 101–108

these critical situations, and hence, it is possible to perform appropriate countermeasures. 4.3 Traffic observer The traffic observer detects critical situations to prevent eventual unrealistic traffic behaviour. Fig. 6 shows that main input/output signals of the traffic observer function.

Fig. 4: Virtual limiting lines that define the front and back sight distances or simulation window. As it will be explained later, traffic vehicles are generated only within the moving window, i.e., the simulation window, of the selected simulator vehicle. However, the other participating simulator drivers are considered, so that they don’t experience unrealistic traffic behaviour. Therefore, the distances between the target simulator vehicle and the other participating simulator vehicles are compared continuously with the set sight distance. If a moving window of a simulator vehicle starts to overlap with the simulation window of the target simulator vehicle, the latter will be adjusted to enclose both simulator vehicles (see Fig. 5). Thereby, traffic vehicles will be simulated within the so-called overall simulation window for both simulator drivers. Adjustment of simulation window applies also if other simulator vehicles approach the target simulator vehicle.

Fig. 6: Traffic observer function and its input/output signals. The function receives information about the target simulator vehicle, the moving windows of all simulator vehicles, traffic vehicles positions, and the maximum length of traffic vehicles. The main task of this function is to continuously monitor which traffic vehicles reside in which moving window(s). Flags with different values are raised for different situations. For instance, a flag is raised if a traffic vehicle resides in the moving window of a non-target simulator vehicle. This information indicates that this traffic vehicle must not be discarded. For example, consider the situation depicted in Fig. 7, where the target simulator vehicle is accelerating and the overall simulation window is being reset. If traffic vehicles reside incidentally within the moving window of the non-target simulator vehicle, these must not be discarded.

Fig. 5: Overall simulation window when a simulator vehicle approaches the target simulator vehicles. Nonetheless, if the simulator vehicle drives away from the target simulator vehicle, the overall simulation window will be reset. In this case, the overall simulation window will be exactly the same as the simulation window of the target simulator vehicle. That is, traffic vehicles will be generated while considering only the target simulator vehicle. Although one target simulator driver is considered for traffic simulation, the dynamic adjustment of simulation window allows all other simulator drivers to experience realistic and smooth traffic behaviour. However, some critical situations can arise through the dynamic adjustment of simulation window. These may lead to unrealistic traffic behaviour, for example, a traffic vehicle that disappears suddenly near a simulator vehicle. Therefore, an observer is required to detect

Fig. 7: Critical situation with traffic vehicles residing in the moving window of a non-target simulator vehicle. Fig. 8 shows another critical situation that the traffic observer detects. A traffic vehicle is leaving the simulation window of the target simulator vehicle, while a non-target simulator vehicle is approaching from the oncoming direction.

106

IFAC AAC 2016 June 19-23, 2016. Norrköping, Sweden

K. Abdelgawad et al. / IFAC-PapersOnLine 49-11 (2016) 101–108

105

window. The function receives information about the simulation window, desired traffic characteristics, current traffic vehicle positions, and eventual critical situations. Fig. 10 shows the input/output signals of the vehicles generation/discarding functions.

Fig. 8: Critical situation with a traffic vehicle residing exactly between the simulation window and a moving window.

Traffic vehicles

Set sight distance

Sum back sight dist. Critical situation flag Set traffic veh. offset speeds

Traffic veh. enable Vehicles generation/discarding Set traffic veh. speeds

Set traffic veh. number Actual traffic veh. positions

Fig. 10: Vehicles generation/discarding function and its input/output signals. The function generates traffic vehicles just beyond the simulation window. In general, traffic vehicles generated from behind are faster than the target simulator vehicle, while those generated in front are slower. Specifically, traffic vehicles are generated with a speed equal to that of the target simulator vehicle +/- set offset speed. To avoid unrealistic traffic queues, the function observes the current positions of simulated traffic vehicles, so that they can be distributed reasonably among road lanes. On the other hand, traffic vehicles traveling slower or faster than the simulator vehicle in the same direction will eventually travel out the simulation window. Similarly, oncoming vehicles will not exist in the simulation window as soon as they pass the sight distance behind the simulator vehicle. While considering eventual critical situation flags of the traffic observer discussed earlier, traffic vehicles are discarded as soon as they leave the simulation window. Fig. 11 shows a general situation for traffic vehicles generation/discarding.

This function adjusts traffic speed and/or density according to the corresponding characteristics provided by the traffic initialisation function (see Fig. 9). Three levels were defined for each of these traffic characteristics: low, medium, and high.

Traffic speed/density controller

Target simulator vehicle Sum front sight dist.

Traffic speed/density controller

4.4 Traffic speed/density controller

Traffic density level

Traffic initialisation

Traffic observer

If the target simulator vehicle stops, the flag is raised to indicate that no traffic vehicles should be generated in front, other than oncoming traffic vehicles. Similarly, if the target simulator vehicle is driving faster than a predefined maximum speed, the flag is raised to indicate that no traffic vehicles should be generated from the back. In all cases, the flag of the traffic observer is forwarded to the vehicles generation/discarding function, discussed later, so that it takes appropriate actions preventing unrealistic traffic behaviour.

Traffic initialisation

Target simulator vehicle

Dynamic windows generation

A flag is raised if the distance between the simulation window and the moving window is equal to or less than the length of the traffic vehicle. This flag indicates that the traffic vehicle must not be discarded; otherwise, the non-target simulator driver would experience unrealistic behaviour.

Traffic speed level

Sim. management station

Set traffic veh. offset speeds Set traffic veh. number

Fig. 9: Traffic speed/density controller and its input/output signals. According to the desired speed and density levels, the function adjusts the offset speed of traffic vehicles and/or the number of traffic vehicles per sight distance behind and in front of the target simulator vehicle. This information is forwarded to the vehicles generation/discarding function discussed in the following sub-section. 4.5 Vehicles generation/discarding Fig. 11: Generation/discarding of traffic vehicles according to the simulation window.

The objective of the vehicles generation/discarding function is to let traffic vehicles exist only within the simulation 107

IFAC AAC 2016 106 June 19-23, 2016. Norrköping, Sweden

K. Abdelgawad et al. / IFAC-PapersOnLine 49-11 (2016) 101–108

The function outputs traffic vehicles speeds together with a signal indicating which traffic vehicles are visible. To guarantee realistic traffic behaviour, traffic vehicles must at least neither collide with each other nor with the simulator vehicles. The following sub-section discusses the driver model that principally prevents vehicle collisions.

following. Traffic vehicles in the free-driving state are unconstrained. Therefore, the driver model forwards the set speeds. For a traffic vehicle in the following state, the driver model adjusts its speed to be equal to its preceding vehicle, so that the safe distance is maintained. Although a driver model with car-following behaviour was implemented in this work, the developed traffic simulation framework can be extended for additional behavioural models, for example, a lane changing model. The following section describes the traffic vehicle module.

4.6 Driver model The driver model controls the behaviour of traffic vehicles according to perceived traffic situations. There are several behavioural models for traffic simulation, like, e.g., carfollowing, speed adaptation, lane changing, overtaking, passing, and oncoming avoidance (Yu et al. (2013)). In this work, the driver model was realised by a car-following behavioural model.

4.7 Traffic vehicles module There are two main methods in traffic research for modelling traffic vehicles: macroscopic and microscopic models. Macroscopic models deal with traffic from an overall or average perspective. They require relatively simple calculations. However, characteristics of traffic vehicles cannot be controlled individually (Nakrachi and Popescu (2010)). On the contrary, microscopic models consider the behaviour of individual vehicles. Yet, a separate model is required for each simulated vehicle (Jaworski et al. (2012)). This increases the computing effort, especially, if a large amount of vehicles is required. The microscopic model was chosen in this work. On the one hand, as traffic is simulated only within a limited area, only a few vehicles are required to achieve reasonable traffic density. On the one hand, human drivers participate in the simulation, where they usually change their behaviour. Therefore, it is necessary to control each traffic vehicle individually according to the position and speed of simulator vehicle(s). Each traffic vehicle is modelled with a simple 1st order lag element together with PI- and PIDcontrollers. Principally, the model can be replicated arbitrarily according to the desired number of traffic vehicles and the available computing capacity. Fig. 13 shows the traffic vehicle model and its main input/output signals.

The car-following model controls the behaviour of a traffic vehicle with respect to a preceding vehicle in the same lane (Wenga and Wu (2001)). A traffic vehicle is considered as following when it is preceded by a vehicle in the same lane, where driving at the desired or current speed will lead to a collision. There are various car-following models in the literature; they utilize different logic and may make different assumptions. For instance, the Gazis-Herman-Rothery models state that the acceleration of the following vehicle is proportional to the speed of the preceding vehicle, the speed difference, and the space headway (Bevrani et al. (2012)). Psycho-physical car-following models use thresholds for the minimum perceived speed difference between the following and preceding vehicles (Schulze and Fliess (1997)). Safetydistance models are based on the assumption that the follower vehicle keeps a safe distance to the vehicle in front (Luo et al. (2011)). A comparison between different car-following models is presented in (Aycin and Benekohal (2014)). For the sake of simplicity, the safety-distance approach is considered in this work. Fig. 12 shows the main input/output signals of the driver model that utilizes a car-following behavioural model. Sim. management station Traffic initialisation Traffic speed/ density controller Vehicles generation/ discarding Traffic vehicles

Traffic vehicle model Controller selection

Simulator veh. positions Simulator veh. speeds

Traffic veh. enable

Controller selection

Set headway distance Set traffic veh. speeds

Switch

Desired speed

Driver model

Set headway distance

Desired traffic veh. speeds

Traffic veh. enable

Preceding veh. pos.

Preceding veh. pos.

+-

PID controller

+-

Actual speed

PI controller

1st order lag element

1 s

Actual position

+-

Fig. 13: Traffic vehicle model and its input/output signals.

Actual traffic veh. positions

If no preceding vehicle is detected by the driver model, the traffic vehicle model receives the set speed, where a PIcontroller is activated to maintain this speed. If a preceding vehicle is detected by the driver model, the traffic vehicle model receives the current position of the detected vehicle, where an additional PID-controller is activated to maintain a safe distance. In both cases, the traffic vehicle model resembles a typical delayed speed behaviour, which is integrated to deduce the position of the traffic vehicle. With appropriate model and controller parameterization, the traffic

Actual traffic veh. speeds

Fig. 12: Driver model and its input/output signals. The driver model receives the current positions and speeds of all simulated traffic vehicles, as well as that of the simulator vehicles. The distances between vehicles are compared with the desired safe distance. According to distance comparisons, a traffic vehicle will be in one of two states: free-driving or 108

IFAC AAC 2016 June 19-23, 2016. Norrköping, Sweden

K. Abdelgawad et al. / IFAC-PapersOnLine 49-11 (2016) 101–108

107

The traffic demonstrator is a graphical interface developed with Matlab. It provides fictive simulator vehicles, where users can adjust their speed and position as desired. Speed and position information is forwarded to the traffic simulation framework, whereas positions of generated traffic vehicles are received by the traffic demonstrator. Consequently, the traffic demonstrator reproduces what happens during simulation runtime, so that traffic behaviour and characteristics can be examined and compared with realworld traffic. The traffic demonstrator provided an efficient test and validation tool during the development of the presented traffic simulation framework.

vehicle shows reasonable smooth movements. Fig. 14 shows the acceleration and speed curves resulting from a typical accelerating maneuver.

6. FUTURE WORK The traffic simulation framework will be developed further to simulate traffic vehicles in the neighbourhood area of each simulator vehicle, i.e., not only within the simulation window of the target driving simulator vehicle. Moreover, other driver models, like, e.g., lane changing, will be added. Considerations for more complex road networks will be taken into account. The developed traffic simulation framework will be integrated within an environment of networked driving simulators. Hence, an expressive evaluation of the generated traffic can be performed with test persons, i.e., interactive simulation. In addition, a method will be developed to compare the computation effort exerted to maintain certain traffic density with and without the proposed approach. Combining the strengths of both microscopic and macroscopic traffic simulation represents another potential extension for the developed traffic simulation framework (Ma et al. (2011)). Thereby, more application scenarios can be addressed while reducing computation effort considerably.

Fig. 14: Acceleration and speed graphs of a traffic vehicle. The traffic vehicle shows a maximum acceleration of 4.4 m/s2, where it reaches 100 km/h in about 18 seconds. However, traffic vehicle models can be parameterized arbitrarily and individually to achieve different acceleration behaviour. 5. FRAMEWORK VALIDATION Several techniques for validating traffic simulation models have been introduced in (Ni et al. (2004)). The traffic simulation framework presented in this work provides good degree of flexibility and scalability, so that traffic behaviour and characteristics can be modified as desired. Hence, quantitative approaches for traffic models validation, using statistical formulas for example, are not applicable in this case. A qualitative approach using graphical displays and animations has been adopted to validate the developed traffic simulation framework (Ni et al. (2004)). To observe the behaviour of the generated traffic, and hence, qualitatively validate the traffic simulation framework, a traffic demonstrator has been developed (see Fig. 15).

REFERENCES Amelunxen, H. (2015). Development of a Closed-Loop Test System for Advanced Driver Assistance Systems. In the Proceedings of the 12th Workshop in Augmented and Virtual Reality in Product Engineering, 245 - 258. Paderborn, Germany. Aycin, M. and Benekohal, R. (2014). Comparison of CarFollowing Models for Simulation. In the Transportation Research Record Journal of the Transportation Research Board, Volume Volume 1678. Behrisch, M., Bieker, L., Erdmann, J., and Krajzewicz, D.. (2011). SUMO – Simulation of Urban MObility. In Proceedings of the 3rd International Conference on Advances in System Simulation, 55 – 60. Spain. Bevrani, K., Chung, E., and Miska, M. (2012). Evaluation of the GHR car-following model for traffic safety studies. In the Proceedings of the 25th ARRB ConferenceShaping the future. Perth, Australia. Bonakdarian, E., Cremer, J., Kearney, J., and Willemsen, P. (1998). ARCHISIM: Generation of Ambient Traffic for Real-Time Driving Simulation. In the Proceedings of the IMAGE Conference, 1 - 10, Scottsdale, Arizona, USA. Dahmann, J. (1999). The High Level Architecture and beyond: technology challenges. In the Proceedings of the 13th Workshop on Parallel and Distributed Simulation, 64 - 70. Atlanta, Georgia, USA.

Fig. 15: Traffic demonstrator for qualitative framework validation.

109

IFAC AAC 2016 108 June 19-23, 2016. Norrköping, Sweden

K. Abdelgawad et al. / IFAC-PapersOnLine 49-11 (2016) 101–108

Espie, S. and Auberlet, J.M. (2007). ARCHISIM: a behavioural multi-actors traffic simulation model for the study of a traffic system including ITS aspects. In the International Journal ITS Research, Volume 5, Issue 1, 7 - 16. Hang, Z., Huang, Y., and Deng, K. (2008). Safety Analysis on Road Sight Distance. In the Proceedings of the IEEE Intelligent Computation Technology and Automation (ICICTA), 461 - 465. Hunan, China. Hang, Z., Huang, Y., and Deng, K. (2008). Safety Analysis on Road Sight Distance. In the Proceedings of the IEEE Intelligent Computation Technology and Automation (ICICTA), 461 - 465. Hunan, China. Jaworski, P., Edwards, T., Burnham, K., and Haas, O. (2012). Microscopic traffic simulation tool for Intelligent Transportation Systems. In the Proceedings of the 15th International IEEE Conference on Intelligent Transportation Systems (ITSC), 552 - 557. Anchorage, Alaska, USA. Kandhai, K., Smith, M., and Kanneh, A. (2011). Immersive driving simulation for driver education and analysis. In the Proceedings of the IEEE 16th International Conference onComputer Games (CGAMES), 288 - 292. Louisville, USA. Lu, N., Cheng, N., Zhang, N., Shen, X., and Mark, J.W. (2014). Connected Vehicles: Solutions and Challenges. In the IEEE Journal of Internet of Things, Volume 1, Issue 4, 289 - 299. Luo, Q., Xun, L., Cao, Z., and Huang, Y. (2011). Simulation analysis and study on car-following safety distance model based on braking process of leading vehicle. In the Proceedings of the 9th IEEE World Congress on Intelligent Control and Automation (WCICA), 740 - 743. Taipei, Taiwan. Ma, J., Sun, J., Li, K., and Zhang, L. (2011). A study on multi-resolution scheme of macroscopic-microscopic traffic simulation model. In the Proceedings of the 14th International IEEE Conference on Intelligent Transportation Systems (ITSC), 1421 - 1426. Washington, DC, USA. Mayenobe, P., Trassoudaine, L., Bellet, T., and TattegrainVeste, H. (2002). Cognitive simulation of driver and cooperative driving assistance. In the IEEE Journal of Intelligent Vehicle, Volume 1, 234 - 239. Miquet, C., Henning, J., and Schick, B. (2010). Continuous Evaluation of Methods for Sensor Data Fusion of VideoBased Driver Assistance Systems, Using Virtual Road Tests. In the Proceedings of the 10th International Symposium on Advanced Vehicle Control, 321 - 327. Loughborough, UK. Muehlbacher, D. (2015). Multi-Driver Simulation – the link between driving simulation and traffic simulation. In the Proceedings of the International Scientific Conference on Mobility and Transport mobil.TUM 2015. Munich, Germany. Nakrachi, A. and Popescu, D. (2010). Modelling and simulation of macroscopic traffic flow : A case study. In the Proceedings of the 18th IEEE Mediterranean Conference on Control & Automation (MED), 1626 1631. Marrakech, Morocco.

Neumann-Cosel, K., Dupuis, M., and Weiss, C. (2009). Virtual Test Drive – Provision of a Consistent Tool-Set for [D,H,S,V]-in-the-Loop, Using Virtual Road Tests. In the Proceedings of the Driving Simulation Conference Europe. Monaco, France. Ni, D., Leonard, J., Guin, A., and Williams, B. (2004). Systematic Approach for Validating Traffic Simulation Models. In the Transportation Research Record Journal of the Transportation Research Board, Volume 1876, 20 - 31. Oeltze, K. and Schießl, C. (2015). Benefits and challenges of multi-driver simulator studies. In the IEEE Journal of Intelligent Transport Systems, Volume 9, Issue 6, 618 625. Olstam, J. (2003). Traffic Generation for the VTI Driving Simulator. In the Proceedings of the Driving Simulator Conference - North America DSC-NA, 321 - 327. Dearborn, Michigan, USA. Olstam, J. (2005a). A model for simulation and generation of surrounding vehicles in driving simulators. Dissertation in Linköpings University. Norrköping, Sweden. Olstam, J. (2005b). Simulation of rural road traffic for driving simulators. In the Proceedings of the 84th Annual meeting of the Transportation Research Board. Washington D.C., USA. Olstam, J. and Lundgren, J. (2008). A framework for simulation of surrounding vehicles in driving simulators. In the ACM Journal Transactions on Modeling and Computer Simulation (TOMACS), Volume 18 Issue 3, 1 24. Pauzie, A. (1994). Human interface of in-vehicle information systems. In the Proceedings of the IEEE Conference on Vehicle Navigation and Information Systems, 35 - 40. Yokohama, Japan. Punzo, V. and Ciuffo, B. (2010). Benefits and challenges of multi-driver simulator studies. In the IEEE Journal of Intelligent Transportation Systems, Volume 12, Issue 2, 354 - 363. Schulze, T. and Fliess, T. (1997). Human interface of invehicle information systems. In the Proceedings of the IEEE Winter Simulation Conference, 1222 - 1229. Atlanta, Georgia, USA. Tideman, M. and van Noort, M. (2013). A simulation tool suite for developing connected vehicle systems. In the Proceedings of the IEEE Symposium on Intelligent Vehicles (IV), 713 - 718. Gold Coast, QLD, Australia. Weng, Y. and Wu, T. (2001). Car-following model of vehicular traffic. In the Proceedings of the International IEEE Conference on Info-tech and Info-net, 101 - 106. Beijing, China. Wright, S., Ward, N., and Cohn, A. (2002). Enhanced presence in driving simulators using autonomous traffic with virtual personalities. In the ACM Journal of Teleoperators and Virtual Environments, Volume 11, Issue 6, 578 - 590. Yu, Y., El Kamel, A., and Gong, G. (2013). Modeling overtaking behavior in virtual reality traffic simulation system. In the Proceedings of the 9th IEEE Asian Control Conference (ASCC), 1 - 6. Istanbul, Turkey.

110