Shi Xueying Sun Zhicheng Zhu Tong (School of Automobile, Chang'an University Key Laboratory of Automotive Transportation Safety Enhancement Technology, Ministry of Transport, Xi'an Shaanxi 710064, China) Abstract: Great differences present between human driving behaviors. It is essential to model the heterogeneous driving behavior affected by varies surrounding traffic environment and collision factors. Based on natural behavior rules, we build a decision-making model of driving behavior where following and lane change are heterogeneous. A series simulations were conducted as well as the vehicle's driving behavior, potential conflict, collision, congestion and other traffic flow evolution phenomena were investigated on urban arterials. According to the operation data of vehicle agent output by simulation, the driver's heterogeneous speed regulation, lane change characteristics, accident impact and space-time distribution of traffic flow are analyzed. Simulation results show that the average speed of aggressive drivers is higher than that of conservative and ordinary drivers due to frequent lane changes under the condition of the same expected speed. In severe congestion, conservative drivers tend to queue orderly, resulting in a faster decrease in their patience value, which makes no significant difference in the timing of lane change motivation of three types of drivers. The more collision vehicles, the more congested the traffic will be, and the average speed of the traffic flow will be lower. After the collision vehicles resume driving, the speed is lower than before. This model reflects the process of traffic flow collision, congestion and dissipation, which has reference significance for understanding the evolution of traffic flow and accident prevention. Keywords: Traffic simulation; accident; heterogeneous driving behavior; rule-based model；multi-agent system

CICTP