Roundabout Vehicle Scheduling Based On Multi-Sensor Fusion Detection and Reinforcement Learning
摘要
Roundabouts are considered to have the effect of reducing vehicle conflicts and improving intersection capacity. However, traffic congestion in the roundabout area has become a long-standing problem. In this article, we try to solve this problem by controlling the density of vehicles in round abouts. First, we established a multi-sensor fusion detection mechanism to detect the vehicle density in the roundabout and its surrounding areas. Then, in order to learn the impact of different areas of the roundabout connection on its vehicle density, we formulate the roundabout problem as a multi-armed bandit problem(MAB) and build an online learning algorithm based on Thompson sampling to solve it. This algorithm can achieve the effect of controlling the vehicle density of the roundabout and reducing traffic congestion by interacting with different areas based on detecting result of vehicle density.