Multi-agent-based Urban Traffic Signal Control System

The urban traffic signal control system is the important part of Intelligent Transport System (ITS). The advanced urban traffic signal control system has an important effect on improving the traffic condition and sufficiently utilizing roadway capacity of the existing road.At present, the traffic flow in a great number of cities of our country is the mixed flow of motor vehicles and non-motor vehicles and foot passengers. The actuality makes the traffic engineers to develop our own urban traffic signal control system, basing on foreign experience and combining the fact of our country. The traditional immovable model is unfit for the mixed traffic flow. The main purpose of this thesis was to advance a new real-time selfadapting urban traffic signal control system. The multi-agent technology, which is a branch of the distributed artificial intelligence, was utilized in this paper. Visual C++ was used to realize the system.In this paper, a new frame of urban traffic signal control system was put forward on the basis of introducing the actuality and the control methods of urban traffic signal control system in our cities, and summarizing multi-agent technology. In the system, one intersection agent was disposed at each intersection, and supervised by the master agent of system. The partial global coordination method was used to program the whole control process. In each intersection agent, the widesence knowledge model was used to realize the control strategy, and the control process of forecasting and corresponding and adjusting basing on induction was used to achieve the control, and the study method of dynamic programming was used to establish the initial control project, and Game Theory was used to realize the correspond with the adjacent intersection agent and the master agent, and the method of direct communication of point to point was used to carry out the communication with the adjacent intersection agent and the mater agent.The technic of Object-orientation was used to realize primarily the system basedon Visual C++ as tool. The control effect of the two control methods that were induction control and corresponding control was simulated. The outcome of simulation showed that the effect of corresponding control was better than the effect of induction control.