Abstract: To study heterogeneity of driving behaviors and its influence on traffic flow characteristics, firstly a questionnaire-based survey was conducted to investigate heterogeneous driving behaviors under influences of surrounding vehicles' speed, traffic flow density, number of stops, collision accidents and other factors. Then according to survey results, a meso-traffic flow model based on multi-agent system was established using NetLogo, and heterogeneous driving behaviors and collision factors were deployed to simulate evolution process of traffic flows. The results show that drivers would adjust driving speed to different degrees influenced by surrounding vehicles' speed, and their average driving speed is related to their own characteristics. Behaviors of changing lanes are determined by traffic flow density and number of stops, and heterogeneity of this behavior makes distribution of vehicles in each lane change in a tidal manner and therefore affects traffic flow stability. It is also found that collision accidents can reduce driving speed and cause traffic bottlenecks.

Key words: heterogeneous driving behavior, collision accident, mesoscopic traffic flow, NetLogo, multi-agent system, traffic simulation