Techincal Research On Intelligent Traffic Signal Timing Optimization Based On Deep Reinforcement Learning

Because of the rapid development of China’s economy,the continuous improvement of people’s income levels,and the accelerating global urbanization process,the number of vehicles in urban road networks has continued to increase on a larger base.In addition,many aspects of daily activities in high-quality life have increased the number of vehicles in the urban road network.The increase in the number of vehicles in the urban road network has led to the gradual seriousness and generalization of traffic congestion,untimely traffic dredging,and traffic accidents at intersections.These phenomena have begun to spread from large and medium-sized cities to small and medium-sized cities,and have also become troubled by cities around the world.The main social problems that restrict economic and social development have increasingly attracted the attention of governments.To solve the urban traffic problem,first solve the traffic congestion problem,and the traffic signal timing is an important means to solve the traffic congestion problem.In view of the key role of traffic signal timing in urban transportation systems,developing a more effective urban traffic signal timing strategy is the fundamental way to solve urban traffic congestion problems.In the traffic timing technology,the traffic signal timing strategy based on traditional Q_learning is an important means to solve the traffic timing problem,but it has a cumbersome Q value table establishment and search,the target Q value is easily overestimated,and can not be long-term.Problems such as memory experience have limited effectiveness in dredging environmental traffic congestion.Depth reinforcement learning itself has a deep network and introduced experience pool,greedy strategy,DOUBLE DQN and other method technologies,which can solve the problems of traditional Q_learning.To this end,we propose a deep reinforcement learning(DQN)strategy to optimize intersection signal timing techniques to reduce the number of vehicles in the transportation system and the average travel time of vehicles passing through the intersection.The experimental results show that the traffic signal timing strategy based on deep reinforcement learning(DQN)is better than the traditional Q_learning strategy for clearing intersection vehicles,which can better clear environmental traffic congestion and improve traffic system efficiency.