Multi agent system for traffic simulation in Vietnam = hệ thống đa tác tử áp dụng cho vấn đề mô phỏng giao thông ở Việt Nam. Luận văn ThS. Công nghệ thông tinL 60 48 01

ĐẠI HỌC QUỐC GIA HÀ NỘI TRƯỜNG ĐẠI HỌC CÔNG NGHỆ Trần Tiến Công MULTI AGENT SYSTEM FOR TRAFFIC SIMULATION IN VIETNAM Ngành: Công nghệ thông tin Chuyên ngành: Khoa học máy tính Mã số: 60.48.01 LUẬN VĂN THẠC SĨ KHOA HỌC MÁY TÍNH NGƯỜI HƯỚNG DẪN KHOA HỌC: PGS TS BÙI THẾ DUY Hà Nội - 2013 ORIGINALITY STATEMENT ‘I hereby declare that this submission is my own work and to the best of my knowledge it contains no materials previously published or written by another person, or substantial proportions of material which have been accepted for the award of any other degree or diploma at University of Engineering and Technology (UET/Coltech) or any other educational institution, except where due acknowledgement is made in the thesis Any contribution made to the research by others, with whom I have worked at UET/Coltech or elsewhere, is explicitly acknowledged in the thesis I also declare that the intellectual content of this thesis is the product of my own work, except to the extent that assistance from others in the project’s design and conception or in style, presentation and linguistic expression is acknowledged.’ Hanoi, October 7th, 2013 Signed i ABSTRACT The VTS is a system that allows users to design different road systems as well as to create different simulation scenarios with different agent profiles It was built in 2010 based on the theory of Agent and Multi Agent System During 2011 and 2012, it was improved and many experiments were performed regarding to the real data collected from VOV traffic online The results are promising and we hope that it could be able to help the traffic planners to solve the sore issues of traffic in Vietnam at the moment PUBLICATION *The Duy Bui, Duc Hai Ngo, Cong Tran, Multi-agent based Simulation of traffic in Vietnam, 13th International Conference, PRIMA, Kolkata, India, pp 636-648, 2010 ii TABLE OF CONTENT List of Figures v List of Tables v List of Abbreviations vi Acknowledgement vii Chapter Introduction 1.1 Motivation, objectives and approach 1.2 Outline of the thesis Chapter Literature Review 2.1 Models of traffic simulation 2.1.1 Scope 2.1.2 Time 2.1.3 Multi-agent system for traffic simulation 11 2.2 Conclusion 14 Chapter Vietnam Traffic Simulator 15 3.1 Introduction to multi-agent system 16 3.1.1 Agent 16 3.1.2 Multi Agent Systems – MAS 18 3.1.3 Agent based model – ABM 19 3.1.4 ABM development 19 3.2 Modeling 20 3.2.1 The road system 21 3.2.2 Agents representing traffic participants 22 3.2.3 Agent’s planning 23 3.3 Improvement 26 3.3.1 Additional Features 26 iii 3.3.2 Correction 27 Chapter Evaluation 29 4.1 Method 29 4.2 Results 34 4.3 Discussion 37 Chapter Conclusion 38 5.1 Conclusion 38 5.2 Future development 38 REFERENCES 39 iv List of Figures Figure VISSIM visual interfaces Figure 2.VISSIM statistic fuction Figure 3.Traffic light simulation system 10 Figure Highway simulation system 12 Figure A part of highway road .13 Figure Road Area 22 Figure Connection road areas 22 Figure Detecting possible collisions 26 Figure Interface and simulation of the traffic light 27 Figure 10 Some examples of real time traffic data 30 Figure 11 the Khuat Duy Tien – Tran Duy Hung crossroad in the simulator .30 Figure 12 the Khuat Duy Tien – Tran Duy Hung crossroad captured by the traffic camera 31 Figure 13 Distribution of inflow vehicles in real data .33 Figure 14 Timegraph of inflow inflow vehicles in real data 33 Figure 15 The worst case of achieved results 34 Figure 16 The best case of achieved results 34 Figure 17 Normalization of achieved results .35 Figure 18 The decrease rate of velocity 36 Figure 19 Traffic light data observation 36 List of Tables Table An example of randomized parameters 28 Table Parameters of KDT – TDH crossroad 30 Table The information query form .32 Table An example of query data 32 Table Default parameters of the simulation 32 v List of Abbreviations MAS – Multi agent system ABM – Agent Based Model VTS – Vietnam Traffic Simulator VISSIM – Visual Traffic Simulation System KDT – Khuat Duy Tien TDH – Tran Duy Hung PH – Pham Hung HL – Hoa Lac vi Acknowledgement First and foremost, I would like to express my deepest gratitude to my supervisor, Ass.Prof Bui The Duy, for his patient guidance and continuous support throughout the years He always appears when I need help, and responds to queries so helpfully and promptly I would like to give my honest appreciation to my co-partner Ngo Duc Hai for his kindly support although he had to prepare for his study oversea I would also like to thank my friend, Vu Tien Thanh, for his kindly help I sincerely acknowledge all my lectures in University of Engineering and Technology, Vietnam National University, Hanoi, for guidance in my master study Finally, this thesis would not have been possible without the support and love of my family Thank you! vii Chapter 1.1 Introduction Motivation, objectives and approach In Vietnam, the transportation system is now facing many problems in terms of congestions and accidents Especially in big cities like Hanoi, the transportation system is chaotic, due to narrow roads, increasing number of vehicles, and lack of consciousness to follow the traffic rules from participants Many solutions have been proposed and implemented which imposed a great effect on the development of the transportation system itself as well as awareness of the whole society However, most of these solutions usually require a huge financial effort to be able to prove effectiveness Therefore, a method which helps reduce the cost of improving the current transportation situation should draw attention of researchers It is the reason why I was motivated to my thesis regarding to this theme In developed countries, transportation planners always have to have a strategic vision which can identify a clear plan to develop the transport system Such knowledge could be attained by experimenting on traffic simulators With information provided by these simulators, the policy makers can figure a way to reduce the cost of traffic infrastructure building Literally, the use of multi agent system in simulating the behavior of the society is a common trend of solving problems like transportation Following this trend, we started to build the Vietnam Traffic Simulator (VTS) based on the multi agent system model under the guidance of Assc Prof Bui The Duy in 2010[17] This thesis mainly aims to strengthen the correctness of the VTS To phrase it another way, the completion of the evaluation for this simulator is the main target of this thesis It requires some approaches in both proactive and reactive ways With the base knowledge acquired from research of MASs, I added some additional features and improved the structure of VTS to be more suited for the traffic in Vietnam To be more specified, I had gathered data from many sources, had added a function, had corrected the behavior of the simulator and then I implemented to evaluation phase 1.2 Outline of the thesis The outline of the thesis is as following: Chapter will be the literature review about traffic simulation models and the approach based on the Multi-agent model Chapter is about some main features of Vietnam Traffic Simulator, including some new improvement after the short paper presented in PRACSYS 2010 [17] The evaluation steps will be presented in chapter The last chapter is the conclusion and future research A’s current speed is higher than its safe speed limit, it will reduce speed Otherwise, it will decide to steer Figure Detecting possible collisions  Determining alternative route to avoid collisions When there might be collisions in the planned route, an alternative route is calculated so that the alternative route is in parallel with the optimal route 3.3 Improvement In the scope of research, the VTS is well built with efficient function tests Therefore, the programme runs smoothly without any troubles in representing the traffic flows However, there are still many functions to be included in this programme such as Traffic Light, input and output data Moreover, there are also some mistakes in the agent based model that need to be corrected In fact, those are not mistakes However the attributes were still not appropriated for the traffic in Vietnam We will describe in details of these improvements in this section 3.3.1 Additional Features  Traffic lights 26 Traffic light is one of the most important parts of transportation system There are many types of traffic light for different types of transport roads such as railroads, land roads or air roads In this simulator, we only implement the traffic light for land roads including states: red light and green light We created a plan for agents to recognize the two states of the traffic lights If there is a red sign ahead, they need to slowdown and the first agents coming to the line of red sign need to stop They can begin to accelerate again if the state of the light changes to green In the next step, we added variables in the object TrafficLight and built up controlled function for it In the end, we designed the interface for this new fuction Figure 10 Interface and simulation of the traffic light 3.3.2 Correction Some corrections have been made to increase the reality of the simulation for the traffic in Vietnam  The perception of drivers There are some mistakes in the perceptions of drivers implemented in the old models According to Sameh El hadouaj and Alexis Drogoul [16], the perception of the drivers is what decides their behaviour on the roads The ages 27 or genders of the drivers have minor effect on their actions For that reason, the categorization of driver’s types is not realistic anymore Therefore, I implemented all the drivers with only one category of attributes However, each driver’s properties should be different from others Consequently, I added random properties in each of these attributes to make them different agents with different properties Parameter Value Maximum velocity 50 km/h (+/- 10) Slow velocity 30 km/h (+/- 5) Accelerate 12 km/h/s (+/- 5) Decelerate 20 km/h/s (+/- 10) Table An example of randomized parameters  Parameter of the drivers During the experiment phase, we realized a fact that the Vietnamese people usually not have the ability to estimate the safety travel well By other means, they usually go without concerning about safe distance between each vehicle For that reason, we made the safe distance now is a random attribute of each vehicle and its range is from meters to zero We also increase the range of the steering arc from 1.20 n/s to 1.6 n/s It allows vehicles to move more freely between the lanes 28 Chapter Evaluation Evaluations of new programs are essential Evaluations are undertaken for a variety of reasons: to judge the worth of ongoing programs and to estimate the usefulness of initiatives; to increase the effectiveness of program management and administration; and to satisfy the accountability requirements of program sponsors Evaluation also may contribute to substantive and methodological social science knowledge In this chapter we will describe how we evaluate the VTS with real time data This part is the most important contribution of this thesis 4.1 Method In order to study the accuracy and realism of the simulation, we have used the VTS simulator to run some experiments After a long duration collecting real data from many different sources such as video captured from mobile phone, video captured from digital camera on the top of high buildings, video captured in the VOV traffic online, stored video of Hanoi Transportation department, we tried to select the most appropriated data for conducting experiments Unfortunately, most of the data we can collect is not feasible for experiments due to the limited of traffic lane, the quality of the video or obstacles blocking the view 29 Figure 11 Some examples of real time traffic data Finally, the experiments have been performed on the Khuat Duy Tien – Tran Duy Hung crossroad with a roundabout This is the most suitable location we could choose for the experiments The parameters of this crossroad is described in the table below: Road name Length Width KDT 50m 12m PH 50m 12m TDH 50m 12m HL 50m 14m Table Parameters of KDT – TDH crossroad Abbreviation: KDT: Khuat Duy Tien, PH: Pham Hung, TDH: Tran Duy Hung, HL: Hoa Lac Figure 12 the Khuat Duy Tien – Tran Duy Hung crossroad in the simulator 30 Figure 13 the Khuat Duy Tien – Tran Duy Hung crossroad captured by the traffic camera Inflow and outflow information was acquired from traffic video clips captured from VOV’s camera broadcasting every day at a.m The statistical data is achieved by counting the vehicles in the video We want to determine if the traffic density of statistic data could match with the data obtained from the simulation In details, if the number of vehicle inflow is equal to the input vehicle of the simulation, then the number of vehicle outflow is also equal to the output vehicle of the simulation However, the statistical data contains only a part of the crossroad and it does not provide any detail of the beginning and the end of the lane road Thus, we only calculation based on the income vehicles of the corners of the camera In this experiment, we count only the numbers of motorbikes labeled as small vehicles, cars and trucks labeled as big vehicles for the reason that other kinds of vehicles such as bicycles rarely appear Because it is hard to count the number of vehicles coming in every second, we decided to count the total number of vehicles passing through an entrance every seconds by set the captured video to be paused every seconds The form for the information acquisition is organized as follow: 31 Type of vehicle From … Density (number of vehicles going to the crossroad per seconds) Avg Velocity (km/h) Turn right Go straight Turn left Turn around Table The information query form Motobikes From KDT Density (number of vehicles going to the crossroad per seconds) Avg Velocity (km/h) Turn right 7 9 4 40 Go straight 6 9 4 45 Turn left 3 1 3 3 35 Turn around 0 0 0 0 0 0 Table An example of query data The parameters of the agents are summarized as below Parameter Value Maximum velocity 50 km/h (+/- 10) Slow velocity 30 km/h (+/- 5) Accelerate 12 km/h/s (+/- 5) Decelerate 20 km/h/s (+/- 10) Plan time 1201 ms Steering angle 1.50 n/s Table Default parameters of the simulation 32 In further details, we considered the bus as a kind of big vehicles though it may not function as a normal truck or car The numbers of inflow vehicles per minute is 423 small vehicles and 79 big vehicles which are distributed in the following graphs Vehicle inflow rate From KDT From TDH From PH From HL Figure 14 Distribution of inflow vehicles in real data The incoming statistics counted are illustrated as the graphs below: Incoming small vehicles per block 3s Incoming big vehicles per block 3s 40 15 20 10 5 Car 11 13 15 17 19 21 Motobike 11 13 15 17 19 21 Figure 15 Timegraph of inflow inflow vehicles in real data 33 4.2 Results As settings above, we compare the outflow rate of statistical data with the finished vehicles out of the crossroad in the simulator The result we achieved is described as following graphs: Chart Title 45 40 Axis Title 35 30 25 Real data 20 Simulator 15 10 5 10111213141516171819202122 Figure 16 The worst case of achieved results Chart Title 45 40 35 Axis Title 30 25 Real data 20 Simulator 15 10 5 10111213141516171819202122 Figure 17 The best case of achieved results 34 We ran the simulator with countenance 20 times While in worst cases the output data is not matched with the real data (Figure 15), one out of five times we running it, the results showed a similar with the real data (Figure 16) In addition, the total number of outgoing vehicle in the simulation is 382 which is rather similar to that in the real life - 398 If we normalization to all the data achieved during evaluation, we could reach a graph of comparison like below Chart Title 45 40 35 Axis Title 30 25 Real data 20 Simulator 15 10 5 10111213141516171819202122 Figure 18 Normalization of achieved results With the same setting, we added some more vehicles to test the limitation of the road capacity As a result, the average travelling velocities of vehicles decreased along with the increase of vehicles participating in With a total of 600 vehicles travelling in this road which is 200 more from the real life traffic, the congestion happens sooner or later We can see the decrease rate of velocities in this graph 35 Velocity 50 40 30 Velocity 20 10 400 425 450 475 500 525 550 575 600 Figure 19 The decrease rate of velocity Another evaluation will be presented in this section is the traffic light simulation However it is really hard to count the number of vehicles stop at the red light So we need to think of another method to collect the feasible data in order to let the evaluation continued Figure 20 Traffic light data observation 36 4.3 Discussion The simulation showed that with the same number of input statistical data, we can achieve the relative outflow rate In addition, through the statistic component report of the simulation, the average velocity of the vehicle after finishing travelling is around 40km/h which is mostly the same as the average of the small vehicle travelling in the real crossroad Another highlighted point is that: the setting of the simulation is randomized Therefore the output data of the simulation is also varied However, the data collected is just one minute of the real life traffic In fact, the traffic can happen just like the situation in the simulation To a degree of extent, we can say that this is an evidence to show that the simulator can be able to perform well on real data The test for capacity of a road is important for the policy makers, however we not have equivalent real data sample to scientifically prove the correctness of the simulator For that reason, it should be left for future research The traffic light evaluation is another point to be discussed, e.g we need a survey to be able to set the number of people who come to opposite lane when waiting for the red light 37 Chapter 5.1 Conclusion Conclusion In this thesis we presented a new approach to solve the traffic problem in Vietnam It results in VTS - a new model of traffic simulation in Vietnam which can help transportation planners to find treatments for many traffic problems such as congestions as well as to save money to test new designs before actually build the transportation infrastructure After gaining experience from researching the state-of-the-art in the traffic simulation and multi agent system, my personal work are joint-created VTS, improved it and conducted experiment to evaluate its correctness With real data gathered from captured video stored by VOV traffic, we could evaluate some main features of the VTS and are able to prove some correctness of the simulator However, there are still many issues left due to the lacking of experiment data We will work for it in the future development presented in the section below 5.2 Future development In the future, we would like to have more data to be able to perform more specific experiments, especially experiments on the traffic light system Without the suitable data, not much work can be done further Besides, there are still many features that need to be improved such as the interface of the system, the method to allow statistical data to be inputted 38 Chapter REFERENCES [1] Adina Magda Florea Introduction to Multi-Agent Systems, In Proc of Continuous Education Program on Intelligent Agents Technology and Knowledge Processing, Bucharest, 2001, pp 49-60 [2] Agent Based Modeling FAQ [3] Charles M Macal & Modeling and Simulation Michael J North Introduction to Agent-based [4] Charles M Macal & Michael J North Tutorial on Agent-Based Modeling and Simulation Part 2: How to model with Agents Proceedings of the 38th conference on Winter simulation, 2006, pp 73-83 [5] Differences between objects and agents [6] FERBER, J., 1999 Multi-agent Systems: Introduction to Distributed Artificial Intelligence England: Addison Wesley [7] Green Light District [8] James J Odell Objects and Agents Compared Journal of Object Technology, vol.1, no.1, 2002, pp 41-53 [9] Katia P Sycara Multiagent Systems AI Magazine, 1998 [10] Matti Pursula, Simulation of Traffic System – An Overview Journal of Geographic Information and Decision Analysis, vol.3, no.1, 1999, pp 1-8 39 [11] Michael Wooldridge, An Introduction to Multiagent System (second edition) John Wiley & Sons, 2009, pp 15-45 [12] Peter Stone and Manuela Velos Multiagent Systems A Survey from a Machine Learning Perspective, Springer Netherlands, 1997, pp 345-383 [13] Rahul Sukthankar, Dean Pomerleau & Charles Thorpe, SHIVA: Simulated Highways for Intelligent Vehicle Algorithms Proceedings of Intelligent Vehicles '95, September, 1995, pp 332-337 [14] Thomas Fotherby Visual Traffic Simulation, 2002 [15] Todd Sundsted An introduction to agents http://www.javaworld.com/javaworld/jw-06-1998/jw-06-howto.html [16] Sameh El hadouaj, Alexis Drogoul, Stéphane Espié How to Combine Reactivity and Anticipation: The Case of Conflicts Resolution in a Simulated Road Traffic Second International Workshop, MABS, pp 82-96, 2000 [17] The Duy Bui, Duc Hai Ngo, Cong Tran, Multi-agent based Simulation of traffic in Vietnam, 13th International Conference, PRIMA, Kolkata, India, pp 636-648, 2010 40