A Multi-Agent Negotiation Approach To Simulate And Analyze Drivers' Route Choice Behavior

Urban Traffic Flow Guidance System (UTFGS) is one of the most important research contents of Intelligent Transportation Systems (ITS). UTFGS takes the measure of guidance to improve traffic system performance, and finally realizes more efficient assignment in the whole road network. Dynamic route choice is the kernel theory of UTFGS. Actual researches on dynamic route choice mostly focus on centralized disposal method, which is always limited in practical application for its complexity and slow calculation speed, while distributed disposal method is another way to solve dynamic route choice problem. In this paper, the problem is consided from the point of distributed disposal, and taking advantage of intelligence and learning ability of agent to help drivers deal with traffic information. Finally negotiation among agents realizes route choice. In the process of negotiation, drivers' travel intention will be dominant, then to enhance network performance, which will give attention both to the assignment principle of user optimal and system optimal.Firstly, the influence factors of drivers' route choice behavior are analysed. By carrying through decision-making intent investigation, the evaluation criteria and their corresponding weights and anchor points that drivers pay attention to during the process of route choice are confirmed. On the basis of survey, the utility functions of each objective and linear utility maximization model used for measuring route performance are set up, which provide gist for agents to select routes.According to multi-agent theory, three types of agents are modeled: system manager agent (Agent-TMS), mediator agent (Agent-ISP) and driver agent (Agent-IRCS), and the route choice negotiation model between Agent-IRCS and Agent-ISP is also built. By using multi-agent simulation software, Starlogo, three different travel scenarios: no information no negotiation travel, offering information but no negotiation travel and both offering information and negotiation travel, are tested. Each scenario is simulated under small (50:150), moderate (100:100) and big (150:50) travel proportion conditions where numerical value is the travel number ratio of common vehicles to experiment vehicles. Furthermore, the best travel proportion of the third senario is studied as well. Simulation results not only proved the significance of traffic information, but also testified the validity and superiority of negotiation.