Research on Coordinated Control Method of Trunk Line in Holographic Traffic Perception Environment
An improved trunk coordinated control model based on multi-agent deep reinforcement learn-ing was proposed in this paper.The real-time traffic information of holographic intersection was mod-eled discretely,and it was used as the state input agent together with traffic characteristic parameters such as traffic delay and parking times.Taking the delay of trunk traffic flow and the number of stops as reward functions,the public period of trunk intersections was gradually adjusted.The phase differ-ence between intersections was updated by time step,and then the adaptive control of trunk intersec-tions was realized.Taking the traditional deep reinforcement learning(DRL)control method and MAXBAND coordinated control method as reference models,the average vehicle delay and average queue length were compared in low peak period,flat peak period and peak period respectively.The re-sults show that the control model can reduce the average vehicle delay by 21.6％,31.8％and 22.1％,and the average queue length by 34.3％,18.4％and 24.1％,which shows that the proposed method can effectively improve the trunk line traffic efficiency in the holographic sensing environment.
NETL
NSTL
万方数据
