Agent Based Cooperative Control Strategy Of Traffic Signal For Urban Intersections

With the rapid development of urbanization,traffic congestion is a crucial problem affecting the quality of life in the modern cities.Instead of changing the existing structure of road network artificially,Intelligent Transportation Control Systems has been widely accepted and developed as the more advanced solution to alleviating urban traffic congestion by improving traffic transportation efficiency.The intersection collaborative control based on multi-agent technology is widely used as a model-free control method.The agent senses the state of the environment to perform control actions without the need for precise control models;and to participate in collaborative control of the intersection without phase and bandwidth requirements the intersections can be randomized.The agent has the ability to self-learning,and they gain feedback on the control effect by sensing changes in the environment to adjust decision-making information,and constantly optimize decision-making.Intersection collaborative control research work in the following difficulties: In the hierarchical collaborative control mode,the implementation of the central agent and the master agent model is complex.Secondly,the cooperative decision state and the spatial dimension of the behavior space are too high,which leads to the difficulty of convergence and fixed cycle phase sequence conditions,green time utilization is not high.In view of the above difficult problems,this article has done the following work:(1)The basic conditions of collaborative control of intersections are studied,and the prerequisites of intersections to participate in collaborative control are defined.The intersecting co-control elements are studied to determine the main controllable parameters;the evaluation index of road network performance is analyzed,and the appropriate evaluation index of intersection control is selected;(2)This paper studies the multi-agent modeling method,analyzes the mechanism of the intersection signal control system,establishes the intersection signal control agent model,and selects the appropriate interactive mode for the multi-intersection signal control agent collaborative decision system;(3)The collaborative decision-making method of intersection signal control agent is studied,and the cooperative decision of multi-intersection is realized based on the multi-agent reinforcement learning foundation.The state and action of the intersection signal control agent are defined,which can reduce the state-action space dimension and speed up the iterative convergence speed so as to find the optimal solution.In this paper,we selected four intersections of Xinnan Road,Du Song Road and Qingfeng South Road and Yinxue Road,which are located in Chongqing Liangjiang New District.The simulation results show that the algorithm control effect is compared with the induction control effect.The results shows that the cooperative control decision method of the intersection signal control agent has a good effect on the traffic flow control of the experimental road network in a variety of traffic conditions,and achieves the purpose of improving the efficiency of road network.Finally,the paper summarizes the shortcomings of the research work,gives the direction of further research.