Cooperative Control Method Of Intersection Signal And Lane In Vehicle-road Cooperative Environment

With the continuous expansion of the scale of urban road network and the continuous growth of the number of all kinds of motor vehicles,traffic congestion has become an important incentive to restrict the efficient and safe operation of modern cities.Intersections are the key nodes of urban traffic system and the basic unit of traffic control.With the support of artificial intelligence,autonomous driving,5G communication and big data technology,the theory,method,technology and practice of mixed traffic flow signal control at urban intersections under vehicle-road collaborative environment are the hot topics concerned by intelligent transportation,intelligent control,computer science,artificial intelligence and other disciplines.It has become an important content and basic consensus for countries all over the world to develop smart cities and intelligent transportation systems to build a collaborative urban transportation environment.In this context,this paper takes the holographic intersection in the vehicle-road collaborative environment as the object.Based on the existing achievements of domestic and foreign scholars,this paper firstly studies the mixed traffic capacity and vehicle ecological driving speed trajectory guidance at the entrance of the intersection in the vehicle-road collaborative environment,and then studies the collaborative control strategy of signal allocation and variable lane setting at the intersection in the vehicle-road collaborative environment.Then,the collaborative control theory and method of signal matching and variable lane setting at vehicle-road collaborative intersection based on machine learning are studied.The main research contents and achievements of this paper are as follows:(1)In terms of the research on the operation rule,traffic mechanism and traffic capacity of mixed traffic flow in a collaborative vehicle-road environment,based on the internal relationship between the stable state of vehicle following and the macroscopic basic diagram of traffic flow and traffic capacity,the calculation formulas for the homogeneous traffic flow of a single vehicle type and the heterogeneous traffic flow capacity of a mixed vehicle at the intersection entrance are established.(2)In terms of the research on the speed trajectory guidance strategy of mixed traffic flow vehicles and teams in the entrance control area of the vehicle-road coordinated intersection,the ecological driving speed trajectory guidance model of vehicles in the entrance control area of the intersection was established respectively under the condition that the current signal was green and red,with the constraint of ensuring that vehicles passed the intersection stop line at a constant speed.(3)In terms of collaborative control of mixed traffic flow at intersections under vehicle-road collaborative control environment,the objective function and constraint conditions of collaborative control optimization of signal allocation and variable lane setting at vehicle-road collaborative intersection are constructed,and the collaborative control strategy and algorithm process of signal allocation and variable lane setting at intersections based on the collaborative control optimization model are given.The effectiveness of the cooperative control strategy is verified by simulation experiments.(4)In the research of traffic signal control at vehicle-road collaborative intersections based on machine learning,the intersection signal control framework based on Nature DQN algorithm is constructed based on the idea and framework of deep reinforcement learning,the vehicle position matrix and speed matrix description of environmental state are given,and the action space of collaborative control at intersections with variable lane setting is designed.The reward function reflecting the change trend of the real traffic state at each sampling time step intersection was defined.The convergence performance of the model algorithm was improved by introducing the dynamic greedy strategy and the empirical replay mechanism.The simulation experiment verified that the traffic signal control method based on deep reinforcement learning was better than the traditional timing control and adaptive control.