Microscopic urban traffic simulation with multi-agent system

Computer traffic simulation is important for making new traffic control and traffic guidance strategies. Microscopic traffic simulators can model traffic flow in a realistic manner. In this paper, a framework of a microscopic urban traffic simulator based on multi-agent system is introduced. The traffic light control agent and the vehicle-driver agent, which are the most important agents, are described in detail. The agents are implemented in a prototype simulator program written in MS C++, which consists of multi-lane roads, intersections, traffic lights and vehicles. Under the environment of the simulator, preliminary experiments have verified the efficiency of a reinforcement learning traffic light control method designed by us.