Study On Optimal Path Problem Of Traffic Network Considering Real-time Travel Time Estimation

The optimal path problem is the core problem of the path guidance system of intelligent transportation system subsystem,and reliable travel time estimation is a prerequisite for route guidance,and it is of practical importance to study the optimal path based on real-time travel time estimation.In this paper,we studied the optimal paths considering real-time travel time estimation from the perspective of travelers.For urban traffic network,we used historical and current data to predict the arrival speed of traffic network nodes in the short-term future by mining the intrinsic variation law of arrival speed of road section nodes.A combined model for real-time travel time estimation was constructed,with the speed trajectory reconstruction model was combined with the prediction of arrival speed of nodes.Based on this model,an optimal path algorithm was designed,which took a small area road network in Nanjing as the example for verification.The details of the study are as follows:(1)The travel time was modeled as a variable,and the velocity-based travel time estimation model is constructed by mining the relationship between velocity and travel time.The classical piecewise truncated quadratic speed trajectory model was improved through considering the travel velocity variation and the continuity of velocity in the time and space domains.(2)Short-term prediction model was incorporated into the travel time estimation model.We modeled the arrival speed of road segment nodes as a time series,and trained the long and short-term memory neural network to learn the temporal patterns of arrival speed sequences of nodes,then took the future 1 hour as the prediction interval,and put the predicted arrival speed of nodes from the long and short-term memory neural network into the piecewise truncated quadratic speed trajectory model,which constituted a combination model for real-time travel time estimation.We designed a comparison test on a real road,which is aimming to evaluate the fitting accuracy of a single piecewise truncated quadratic speed trajectory model and the combined model for real-time travel time estimation.The experimental results show that the fitting accuracy of the combined model for real-time travel time estimation is improved than that of the single piecewise truncated quadratic speed trajectory model,and the error indicators of combined model are reduced by 46.68%,49.76% and 52.34% on average in the morning peak,midday and evening peak hours,respectively.(3)We took the needs of traveler’s perspective into consideration,A model of the optimal path,which has the least travel time cost,was proposed,and an optimal path algorithm combining Yen’s K shortest paths(Yen’s KSP)algorithm and the combined model for real-time travel time estimation was proposed to solve the model.The travel time of the K shortest paths at estimation,and finally,taking time cost as the selection strategy,the path that takes the least time was selected as the optimal path.For verifying the feasibility and superiority of the real-time travel time optimal path method proposed in this paper,an empirical test was carried out for a small-scale road network in Nanjing,the K-optimal path considering real-time travel time was in comparison with the minimal time path and the optimal path of the classical piecewise truncated quadratic speed trajectory model under a priori decision,The test results show that during peak hours the result of the K-optimal path based on real-time travel time estimation is the same as the minimal time path,and the absolute percentage error of time between them is 3.92%,and the accuracy is improved over the a priori optimal path;the length and elapsed time of the K-optimal path based on real-time travel time estimation are not significantly different from the minimal time path during flat peaks,moreover,the path time and length are reduced by 0.93 min and 58 m,respectively,which are better than the a priori optimal path.We studied the optimal path of real-time travel time under urban traffic networks,and carried out research work in terms of travel time theoretical model construction and solution,construction and algorithm programming of short-term prediction model for arrival speed of node,as well as modeling and solution of optimal path.Through the verification of the example study,the following conclusions are drawn: the optimal path results based on the algorithm of real-time travel time estimation meet the actual road conditions and the results are closer to the real values.This study provided a theoretical basis for the optimal path method for traffic networks.