Evaluation of metropolitan traffic flow with agent-based traffic simulator and approximated vehicle behavior model near intersections

Abstract

In this paper, we introduce a metropolitan traffic simulation with microscopic vehicle agents with approximated behavior near intersections. We simulate a metropolitan traffic flow for Tokyo and surrounding four prefectures with fine-grained traffic demand obtained from Tokyo Person Trip survey. Though this simulator has an ability to manage signal control, it is difficult to obtain the real signal data of the city. Without signal data, the behavior of vehicles in a city becomes too smooth because they do not stop at intersections. This causes differences in traffic volume distribution. In this paper, we introduce a virtual vehicle that appears virtually in the front of the lead vehicle on each road to achieve natural deceleration with a car following speed model. We evaluate the aggregated effect of the virtual vehicle by comparing simulated traffic volume and trip length with real traffic data including road traffic census and person trip survey data.