Traffic Jam Prediction And Optimal Path Planning Based On Tensor Decomposition

With the expansion of the city scale and the dramatic increase in the number of private cars,traffic jam has become more frequent in people's daily lives.Traffic jam not only delays time,but also causes great damage and waste to the environment and social resources.Accurate prediction of location,time,and reasonable route planning for traffic jam can help improve people's travel efficiency and save costs.Therefore,it is necessary and significant to study traffic jam prediction and provide reasonable trip planning.In recent years,traffic informatization has continued to develop,and traffic management departments have accumulated a large amount of traffic flow data,which provides a possibility for prediction traffic jam.The accuracy of traffic jam prediction and the rationality of the optimal path planning have a lot to do with the quality of data.However,due to the inevitable noise or lack of data in the collection and transmission process,traffic jam prediction and path planning are extremely difficult.This paper fully considers the multi-model correlation of traffic data flow in the past,present,future,and builds historical data into tensor form.Through the tensor decomposition method,the missing values in the traffic data are complemented and the average travel time of the vehicles in each road segment is predicted,and then the travel route planning is performed.The main work of this paper includes:(1)Missing data completion.First of all,based on the multi-dimensional characteristics of traffic flow data,the tensor is constructed from historical data,and then the missing values in the data are complemented by means of tensor decomposition.The main purpose is to make full use of traffic flow data in minutes,days,and weeks of correlation.(2)Prediction of vehicles` travel time.Based on the completed traffic data,the historical data and the data to be predicted are used to build a tensor,and we use tensor decomposition method to predict the average travel time of each road segment.(3)Travel route planning.Based on the predicted average traffic time of each segment,calculate the average speed of vehicles on each segment,determine the traffic status of each segment,and then plan the route based on the average speed of the vehicle.(4)Based on the real data provided by Alibaba Cloud's “Smart Transportation Competition”,this paper conducts experiments on the proposed method and verifies the effects of tensor completion and tensor decomposition prediction.The experimental results show that tensor completion helps to improve the accuracy of prediction,and the accuracy of tensor decomposition prediction is better than the traditional grayscale model.In order to visualize the effect of path planning,this paper uses the topological relations between road segments to draw the road network and draw the path planning results on the road network.