Research On Vehicle Driving Path Restoration Algorithms Based On Navigation Trajectories

With the continuous growth of China’s expressway operating mileage,the corresponding operating costs have also increased significantly.Among them,the establishment and maintenance of CPC toll stations and ETC gantry have become an important part of expressway operating costs.Therefore,how to reduce the operational burden,improve the vehicle mileage charging method,and make the vehicle mileage calculation and charging faster and lower cost,has become a key problem to be solved urgently by the high-speed operation department.With the completion of the Beidou series of satellite networks in recent years,the improvement of the hardware level has made the collection accuracy of vehicle location information continue to improve,which makes it possible to use the time series algorithm to restore the vehicle’s driving trajectory and assist in charging.The task of trajectory restoration is to restore the driving trajectory of the vehicle into a gantry sequence,which has become an important research task in the current field of road traffic.First,the existing vehicle trajectory restoration tasks are limited by the trajectory sampling accuracy and over-reliance on highway network information,and the complex spatiotemporal relationships in the trajectory restoration data are difficult to accurately model.It is proposed to apply the deep learning model to the vehicle trajectory recovery task,and a seq2seq-based trajectory recovery gantry sequence(Map Gantry Trajectory Recovery by Seq2 Seq,MGTR)model is designed,and an advanced spatiotemporal sequence structure is used to model the spatiotemporal data.The structure establishes dependencies from the time dimension,so that the model structure can contain both time and space information.The model sets a new map representation method to describe the latitude and longitude,and describes the vehicle trajectory information in combination with the vehicle’s driving direction angle and other related information.At the same time,a temporal component based on a recurrent neural network structure is designed to capture the pre-and post-order dependencies existing in the time series.Finally,input the result of the component output into the output layer to reduce the dimension,and obtain the result of the vehicle passing gantry sequence.Secondly,in view of the existence of data noise in the trajectory data,insufficient utilization of the spatial correlation of the gantry,and failure to consider the weight difference of different trajectory points in the entire trajectory sequence,a number of improvement measures are designed,and a spatiotemporal similarity-based method is proposed.Trajectory recovery gantry sequence(Spatial-Temporal Similarity Map Gantry Trajectory Recovery by Seq2 Seq,STSMGTR)model.STSMGTR designs a novel historically similar trajectory fusion structure,based on the original trajectory sequence,to reduce the impact of data noise by fusing historically similar trajectories.At the same time,the time series model is further improved,and attention mechanism is introduced to increase the proportion of trajectory points that have a greater impact on the restoration of the gantry.In addition,through pre-training,the topological information of the gantry is fully utilized to capture the correlation between the gantry,and the gantry information can be embedded and represented more accurately.Finally,based on the supervision data set of a high-speed road section in Guangdong Province,a comparative experiment and result analysis are carried out with the current mainstream vehicle trajectory restoration algorithm.The experimental results show that the performance of the MGTR model is better than other existing methods,and the prediction performance of the improved STSMGTR model has been further improved compared with the MGTR model.The model proposed in this paper can effectively mine the spatiotemporal correlation in the vehicle trajectory data,effectively restore the vehicle gantry,meet the basic requirements of the gantry restoration,and provide new methods and ideas for the research on vehicle trajectory restoration.