Design And Implementation Of Traffic Signal Optimization Control Method Based On Fuzzy Logic And Reinforcement Learning

In recent years,the rapid development of social economy and the gradual expansion of the scale of cities have been accompanied by a rapid increase in the number of motor vehicles in the city.These have caused a series of problems such as vehicle congestion,frequent accidents,and increased vehicle exhaust emissions,which pollute the environment.The key to solving these problems is to improve the traffic capacity of the road network.By reducing the delay time of vehicles at the intersection,the traffic efficiency of the urban road network can be improved.Under the above background,this paper designs the following three methods to solve the traffic signal control problem of urban single intersections,and conducts experimental comparisons through the secondary development of Sumo simulation software.(1)A fuzzy control method for traffic light signals is designed and realized.The method is based on four-phase phasing sequence to control traffic lights at a single intersection and adopts a two-layer fuzzy control system.The input of the first-level fuzzy control system is the number of vehicles in line and the arrival rate of vehicles,and the output is the current phase and the next phase traffic flow intensity.The second-level fuzzy control system takes the traffic flow intensity of the two phases as input,and outputs the green light extension time of the current green light phase.The final experimental results show that the control performance is better than Sumo’s own control program and traditional fuzzy control method.(2)Use genetic algorithm to optimize the fuzzy control system.The fuzzy rules and membership function parameters are simultaneously coded into chromosomes.The fitness function is designed with indicators such as the average waiting time of the vehicle.In the iterative process,the population decodes the individuals one by one to generate the corresponding fuzzy control system parameters,and then simulates in the simulation software to obtain the individual evaluation results.In the selection process,an elite retention strategy is added to ensure that the optimal individual is not destroyed.After optimization by genetic algorithm,the control performance of the fuzzy control system in controlling traffic lights has been significantly improved.(3)Reinforcement learning method is used to solve the traffic signal control problem of single intersection.This method makes full use of the traffic flow parameters transmitted by the road network.Intercept the road network as a position and velocity matrix,and use it as an input state.When defining the reward,the average waiting time of the vehicle is used as an indicator,and the duration of the phase is output as the action.In order to solve the the overoptimistic value estimation problem,DOUBLE DUELING DEEP Q NETWORK(3DQN)algorithm is used.The final experimental results show that the control performance of the reinforcement learning method is better than Sumo’s own control program.(4)Based on the above algorithm,a traffic signal control system is designed and implemented.This system can decide the signal phase cycle based on the traffic data of a single intersection.The system includes user management,traffic element management,and traffic signal program simulation task management.