Research On Intelligent Control Of Traffic Light Based On Reinforcement Learning

The continuous increase in traffic volume leads to traffic congestion,and it is a serious problem that needs to be solved in the current traffic research field.Improvements in road infrastructure have not been able to solve this problem due to financial and space resource constraints.Traffic congestion not only wastes people's travel time,affects comfort,but also causes the car to emit more exhaust gas,thus polluting the environment.Therefore,it is extremely urgent to solve the adverse effects of traffic congestion on the city.The intersection is the intersection of roads and vehicles,and it is also the place where traffic congestion is the most serious.The traffic lights deployed at intersections direct the traffic flow through the intersections in an orderly manner,so the timing scheme of the traffic lights is an important factor affecting traffic conditions.Therefore,timely adjustment of traffic signal timing scheme according to real-time traffic flow at the intersection can greatly reduce vehicle delay time,improve travel comfort,and alleviate traffic congestion.Artificial Intelligence has received much attention for its ability and potential to handle complex tasks,it can handle large amounts of data and intelligently control it in complex situations.This paper proposes a FRTL control model for the traffic congestion at intersections and the use of Q-learning in reinforcement learning on the fog platform.The model is a distributed multi-agent intelligent traffic signal control method,which treats each intersection as an agent.Since the fog nodes deployed at each intersection share the traffic information of the intersection,the FRTL model not only considers the traffic flow information of its own intersections,but also considers the traffic flow information of the adjacent intersections.The main ideas of the FRTL model are:the fog node sends the collected real-time traffic flow information to the traffic light control module,and the module uses the traffic light control algorithm to calculate the traffic light timing;the fog node will upload traffic flow information and timing scheme to the regional fog server to realize regional traffic information sharing on the fog platform.By combining fog computing and reinforcement learning,the purpose of intelligent coordinated control of traffic lights can be realized,thereby improving road traffic capacity of road network.This paper uses an integrated simulation platform.The platform is built using three softwares:VIS SIM,Excel VBA and MATLAB.By comparing various performance indicators of FRTL control methods,traditional time-segment control methods,and main road control methods.The simulation results show that the FRTL control mode can properly control the traffic light time according to the real-time traffic flow,which can effectively improve the adjustment efficiency and throughput of the intersection,reduce the queue length and waiting time of the vehicle at the intersection,and achieve the goal of alleviating traffic congestion.