Research And Realization Of Vehicle Cooperation Control Method For Large-Scale Unsignalized Intersection

The development of 5G communication and artificial intelligence has promoted the development of intelligent transportation systems.intelligent transportation systems will change the pattern of road safety and traffic management,especially at intersections without traffic signal control,that is,unsignalized intersections.At present,the research on the cooperative control of vehicles at unsignalized intersections is divided into two categories:traditional mathematical models and artificial intelligence reinforcement learning(RL)methods.However,the current research on vehicle cooperative control at unsignalized intersections is still mainly limited to the intersection scenarios with single-vehicle intelligence or a fixed number of vehicles,and there are few studies on vehicle cooperative control at large-scale intersections.In this paper,based on RL,the research is carried out from two aspects of the single-intersection dynamic hightraffic vehicle cooperative control optimization and the multi-intersection vehicle cooperative control.In terms of traffic optimization at a single intersection,this paper proposes a progressive value-expectation Estimation Multi-agent Cooperative Control(PVE-MCC)algorithm based on reinforcement learning.The algorithm designs a value expectation estimation strategy based on incremental learning to avoid the strategy network falling into the local optimal solution and combines this strategy with the generalized advantage estimation algorithm to improve the convergence accuracy and stability of the algorithm.Secondly,the PVE-MCC algorithm designs a multi-objective reward function and a heuristic intervention strategy to improve the comprehensive performance of cooperative control.The research results have been published in the journal "Transportation Information and Engineering".In the aspect of multi-intersection coordination,this paper proposes a multi-intersection Vehicular Cooperative Control(MiVeCC)method.First,this paper proposes an end-edge-cloud distributed fusion computing framework,which realizes vertical and horizontal cooperation between vehicles in large-scale unsignalized intersections.In this framework,a multi-vehicle state representation method and a safety-oriented value representation method are designed to realize the abstraction of the intersection state and the quantification of the vehicle action value.Finally,the multi-layer fusion control strategy is constructed to realize the efficient cooperative control of vehicles at large-scale unsignalized intersections.The research results of this part have been published in the journal "IEEE Transactions on vehicular technology".Finally,this paper builds a simulation experiment platform for unsignalized intersection scenarios for function and performance verification.The experimental results show that the single-intersection PVE-MCC algorithm improves the traffic flow by 30.47%,the bicycle efficiency by 6.25%,and the comfort by 53.82%compared with the existing scheme.The multi-intersection MiVeCC algorithm can improve the travel efficiency of multi-intersection by 4.59 times.The effectiveness of the proposed schemes in this paper is verified.