Research On Traffic Light Area Control Method Based On Deep Reinforcement Learning

In recent years,with the steady progress of China’s modernization drive,the urbanization process has been accelerating,the number of motor vehicles has increased significantly,and traffic congestion has become a widespread problem in major cities.With the development of Internet technology,artificial intelligence,big data,deep reinforcement learning and other technologies have become the research hotspot,related technologies are widely used in the field of traffic light control.In order to solve this problem,we need to find a more reasonable algorithm of traffic light control.The search for more reasonable traffic light control algorithms provides new technical support for solving traffic congestion problems.Because of the complicated traffic flow characteristics and the large size of the road network,it is inevitable to divide the road network into several independent sub-regions for control.The accurate traffic prediction model can better predict the next traffic conditions,which is of great significance for combining the current and future traffic conditions to develop a reasonable traffic light control algorithm.This paper optimizes the shortcomings of the existing traffic light control algorithm.Combined with sub zone division and flow prediction,a traffic light area control method based on deep reinforcement learning is proposed.The main research contents are as follows:Firstly,the role of sub-division in traffic signal control is explored and studied.Aiming at the problem that the traditional subdivision algorithm only considers the static attributes of the road,the traditional Newman fast algorithm is improved by comprehensively considering traffic flow,road length,traffic flow density,queuing length and signal period,the correlation degree between intersections is calculated.The dynamic division of road network is realized.The experimental results show that the dynamic control subarea division method based on the improved Newman fast algorithm is better than the traditional Newman fast algorithm,and can achieve the dynamic subarea division,and the results are more consistent with the characteristics of traffic flow.Secondly,the traffic flow prediction method based on neural network in the existing related research does not consider the spatial characteristics of the traffic flow and the problem of high computational complexity.By combining the CNN network with the GRU network,a new traffic flow prediction model is established.First,the CNN network is used to extract the spatial features of the traffic flow,then the GRU network is used to extract various temporal features of the traffic flow,and finally the prediction results are fused.Experiments show that the traffic flow prediction algorithm based on CNN and GRU network is superior to the traditional LSTM prediction algorithm and can improve the prediction accuracy.Finally,the existing signal control algorithm does not solve the problem of multi-intersection cooperative control,and the control strategy has a certain lag.This paper presents a double DQN traffic light area control method combined with flow prediction.The traffic zone division algorithm is used to determine the area range of traffic light collaborative control,and the traffic flow prediction algorithm is used to improve the forward-looking control strategy.Based on double DQN algorithm,the collaborative control of multiple intersections is realized.Simulation experiments show that the algorithm can reduce the waiting time of vehicles in light traffic flow and heavy traffic flow environment.Improve the traffic efficiency of the road network and ensure the smooth flow of the road network.