Research And Design Of Traffic Situation Prediction And Equilibrium Algorithm

With the development of city traffic and the increasing number of cars,the congestion problem in large city becoming more and more serious.On the one hand,the congestion problem is caused by the increasing demand of travel object and some structure of roads which are unreasonable,on the other hand,the congestion problem is caused by the subjective preference of travel object when they are choosing the travel path,the subjective preference may result in an imbalanced load of road resources,which aggravate congestion finally.The development of smart traffic and smart city will solve the congestion problem,by linking every car in city,we have the way of getting the full-scale traffic data,and solving the problem.The preference of travel object will result in local congestion problem,to solve the local congestion problem,we need a method to evaluate the city traffic state from the macro point of view,then give the guidance to the city travel object from the micro point of view,that is,based on full comprehension of traffic state,law and property solving the congestion problem.First of all,this paper utilizes the traffic graph model and statistical analysis methods to analyze,describe and evaluate the traffic state from a macro point of view.Then,depending on the dynamic spatio-temporal correlation characteristics of the traffic situation,models spatio-temporal correlation,realize the algorithm of predicting traffic situation which means find the transformation law of urban traffic situation.Finally,an adaptive guidance mechanism based on regional potential difference is designed,it decides the intensity of every guidance action,traffic guidance is proposed based on reinforcement learning and fusion of traffic situation prediction,to perform the traffic guidance,it tries to find the balance state between the global traffic state and the local travel object demand,to achieve the traffic situation balance.This paper first summarizes related works and technologies.Then analyzes and evaluates the traffic situation,establishes a traffic situation prediction model based on the attention mechanism.Finally,propose a traffic situation equilibrium model which is traffic situation awareness based on deep reinforcement learning,by the fusion of traffic situation prediction algorithm.In this paper,simulation experiments are carried out,and the experimental results show the algorithm proposed in this paper is better than other algorithms,in terms of stability,execution efficiency and travel time costs.