MULTI-INTERSECTION TRAFFIC CONGESTION CONTROL METHOD BASED ON REINFORCEMENT LEARNING

Abstract

Multi-intersection traffic congestion control is always affected by two problems. The first one is multi-intersection will lead to multi-system status. The second one is the influences between adjacent intersections. Under the influence of the two problems, it is hard to control traffic congestion, therefore, in this paper, we propose the bilayer control model, the upper layer can control the road network area and the lower layer can control the area intersection in parallel in this model. Through combining upper and lower system control, the model can solve the problem of multi-system state effectively. At the same time, the model uses a message-passing mechanism to solve the problem of influences between intersections when the intersection is related to the direct connected intersection. Experiments show the bilayer control model can solve the problem of multi-intersections traffic congestion control effectively.