Research On Collaborative Optimization Method Of Urban Road Traffic Arteries

With the improvement of people’s living standard in our country,the demand for transportation increases,so leading to the traffic congestion problem is becomes more and more serious.Traditional timing control and induction control methods can not adapt to the complicated and changeable traffic environment efficiently.Intersections are the main place of traffic congestion,and there is a great correlation between adjacent intersections.It is an inevitable trend of future traffic development to establish an adaptive intelligent traffic control system to reasonably optimize and configure the signal lights at each intersection and guide the vehicles on the main road to pass effectively.According to the characteristics of traffic flow uncertainty and randomness,the paper first studies the single intersection timing optimization model based on the Sarsa(λ)learning algorithm,and takes the minimum average delay time at the intersection as the optimization goal.Carry out real-time monitoring and update the timing plan based on the monitored information.After the algorithm training converges,the optimal timing plan under various traffic flow states is obtained.In order to solve the problems of Sarsa(λ)learning algorithm storage and large amount of calculation,this paper proposes to use neural network to store the Q value function,and conduct an in-depth study on the BP-Sarsa(A)learning-based single-intersection timing optimization model.As the input of the neural network,the network weights corresponding to the selected timing parameters are updated according to the traffic state and timing parameters to realize the estimation of the unknown traffic flow state.Finally,a single intersection simulation scenario and experimental platform were constructed.Analysis of the experimental results shows that the delay index based on the BP-Sarsa(λ)learning model is better than the Sarsa(λ)learning model,and the BP-Sarsa(λ)learning algorithm can effectively improve Traffic flow efficiency,reduce delay time,reduce vehicle queue time.By analyzing the distance and relevance between adjacent intersections in the main line,and considering the start and stop time of vehicles,the paper proposes a period calculation method considering the interval of the intersection,optimizes the public period of the intersection,and combines the Webster method with the period.Combining optimization methods to get the best common cycle.Introduce the traffic flow ratio coefficient,design the bandwidth of the coordination direction of the main line,and make improvements based on the maximum green wave model method to solve the ideal phase difference.Research on the optimization model of the traffic arterial line based on the BP-Sarsa(λ)learning algorithm,determine the optimal phase difference range according to the ideal phase difference,take the minimum average delay time of the arterial line as the goal,update the phase difference plan in real time,and finally obtain the various traffic flow conditions The best phase difference scheme.Then,the simulation scenario design of the trunk line is carried out,and the average delay time and queue length of the coordination direction and the non-coordination direction of the trunk line are compared and analyzed.The experimental results show that the traffic trunk optimization algorithm based on BP-Sarsa(A)learning can effectively reduce the average delay time and queue length in the coordination direction.To improve the problem of urban traffic congestion.This paper aims to reduce the average delay time in the trunk line network,and establishes the trunk line collaborative optimization system.This method can effectively reduce the average delay of vehicles in the coordinated direction of the trunk line,and achieve the purpose of improving the traffic condition.The experimental results show that the optimization effect of this method reaches the expected goal.