Study On Urban Traffic Signal Control Algorithm Based On Agent

Urban traffic control as a big complex system, makes that the traditional control methods based on exact mathematic model can't meet the increasing society demands, and forces the people to utilize the idea associated with intelligent control in traffic control. In resent years, intelligent Agent and Multi-Agent technology were becoming the top research subjects gradually in the intelligent traffic control.In this paper, the main works are aiming at the intelligent Agent control in urban traffic signal intelligent control methods, the works involve following several aspects:1 Based on Agent technology and Q-learning theory, we propose an Agent traffic controller and realized the control of a real single intersection in Beijing. In this paper, the learning function of Agent is realized by Q-learning which combines the rules based on knowledge. A reward function for real traffic condition is proposed to control traffic dynamically and on real-time. The performance of this method is evaluated by the traffic simulation software Paramics, experimental results indicate the effect of the approach is better.2 In this part, an approach based on multi-Agent technology is applied to urban traffic trunk roads coordination control. Firstly, the urban traffic trunk roads are regarded as a large scale system, and the subsystems are the intersections in the trunk road, on the basis of it the description and function of the decision Agent and intersection Agent are given. Then the intersection Agent's control function based on the distributed Q-learning is described. Finally some simulations on a road net including two intersections are done by Paramics traffic simulation software, and the simulation results indicated the effectiveness of the approach.3 On the bases of the Multi-Agent technology and the two parts above, we do some studies on the control of urban area traffic. In this part, the urban area traffic are regarded as a large scale system, and the subsystems are the intersections in the area. Based on decomposition-coordination idea of large scale system theory, each intersection has its own control Agent which manages the green phase switching time and switching sequence dynamically according to its own and its neighbor's traffic situations, in order to make the whole area traffic smoothly. Intersection Agent controller is composed by two parts: study control module and phase sequence chosen module, and the two parts work coordinately. Finally some simulations on a real urban area including nine intersections are done by Paramics traffic simulation software, and the simulation results indicated the effectiveness of the approach.Finally, we summarize the researching contest of the paper, and put forward some suggestions to the further research in these fields.