Research On Traffic Optimization Based On Reinforcement Learnin

With the development of society and economy,private car ownership in China is increasing,so the traffic congestion problem is increasingly highlighted,traffic jams will not only to the impact of road traffic,idling vehicles will be massive emissions of polluting gases,environmental pollution,at the same time,the road congestion caused the driver be agitated mood,so that traffic accidents are more likely to happen.Therefore,traffic congestion has become a key problem urgently needed to be solved in the field of road traffic.As the most important part of the road traffic,the intersection plays a reasonable role in diverting the traffic flow in all directions,so that the traffic flow in all directions runs in the most appropriate way,so as to reduce traffic congestion and traffic accidents.However,the intersection is also one of the most prone to traffic congestion in road traffic.The signal lamp deployed at the intersection has the function of controlling the traffic flow in all directions and making it pass the intersection orderly.Therefore,in a variety of ways to reduce traffic congestion,many methods are to control the intersection signal light to adjust,set reasonable traffic signal timing scheme,so as to reduce traffic congestion and increase the comfort of travel.With communication technology in recent years,the development of information technology,such as artificial intelligence technology,many new methods dealing with the problem of traffic congestion,such as artificial intelligence technology in the reinforcement learning algorithm,and reinforcement learning through continuous interaction with the environment,learn by trial and error,and achieve the level of intelligent decision-making,therefore,many scholars applying reinforcement learning in the field of intelligent transportation,The reinforcement learning algorithm is used to control traffic signal lights,and the obvious effect is achieved.At the same time,the emergence of 5G technology makes automatic driving possible,and the low delay of5 G technology ensures the real-time transmission of information.Therefore,in the future,autonomous driving vehicles and manual driving vehicles will appear on the road at the same time,and there are few studies on the road traffic mode in this case.The main work of this paper is as follows:Built a simulation platform based on traffic simulation software(Simulation of Urban MObility,SUMO),and set the corresponding road environmental parameters to simulate the real road traffic state,and carried out simulation experiments on the two models designed in this paper,and analyzed the simulation results in detail.The depth of the two kinds of reinforcement learning model to control the traffic light at the intersection,the first model using convolution neural network for road status information as input,by reinforcement learning model,signal control scheme,and optimize the traffic conditions,the results showed that the average waiting time of the vehicle was reduced by about 78% compared to the traditional signal control system;The second model uses the number of vehicles queuing on the road as input,and at the same time combines the vehicle control system responding to the signal to regulate the autonomous vehicles,and tests the performance under different proportions of autonomous vehicles.The simulation results show that the model can significantly reduce the average waiting time of vehicles.At the same time,when the proportion of autonomous vehicles is 50%,the average waiting time of vehicles in the case of mixed vehicles can be significantly reduced.When the proportion of autonomous vehicles is greater than 50%,the average waiting time of vehicles decreases slowly with the increase of the proportion of autonomous vehicles.