Rearsch And Application Of Road Network Travel Time Estimation Based On GPS Data

With the rapid development of China’s economy in recent years,people’s travel has become more convenient.It is a very common choice to choose self-driving,taxi and other cars when traveling in the city.Not only that,almost all cars are equipped with GPS devices,or GPS data can be collected using mobile phones,which provides considerable convenience for data analysis.The collected data contains time,position,direction,speed and other information.Analysis of these data can effectively empower traditional transportation planning.At the same time as the convenience of automobile travel,it also puts a lot of pressure on the urban road network.A large number of roads are congested due to excessive traffic during peak hours.In this scenario,if you can estimate the condition of each road in advance and calculate the possible transit time of the road,you can make better planning for road traffic.And how to use the GPS data collected by the vehicle to estimate the transit time of the road.Traditional road traffic status data mainly comes from various sensors installed around the road,which will inevitably produce data dead ends,resulting in only the analysis and monitoring of hot roads.Using GPS data to estimate road traffic has high practical value and research significance.Analysis based on a large number of vehicle GPS data can improve the analysis of road conditions from multiple aspects: no need to arrange a large number of sensor equipment to save costs And later maintenance costs,data loss due to sensor failures and data blind spots where sensors cannot be collected can also be analyzed and estimated accordingly.This article focuses on the processing and analysis of this GPS data.The main contributions include:(1)A new road matching algorithm is proposed.Road matching is the process of associating GPS coordinate points with roads on the road network.The road matching data obtains the road location information that the original data does not have,so that further research and analysis can be carried out.In this paper,a new road matching method FWMM(Fast Weight-based Map Matching)is proposed,which can run stably on data with variable sampling rate.FWMM is a global road matching method that considers the position,direction,and transfer relationship between GPS points and roads,and gives a global optimal matching result.In addition,in order to reduce the computational cost of the global algorithm,FWMM has an adaptive acceleration matching strategy.Experiments show that this strategy can effectively accelerate the matching process and greatly reduce the average matching time.(2)A method for estimating road transit time based on road matching results is proposed.Since this method is based on the road matching results of GPS trajectories,the transit time of roads not covered by the sensor can be estimated.This method first analyzes the results of the road network matching,extracts useful data for integration and collection,divides the data into their respective time windows based on the sampling time of the data,and makes a preliminary estimate of transit time;then,the tensor decomposition method is used Completion of missing data;In order to improve the accuracy of the method’s estimation results and be compatible with existing sensor data,a correction scheme for model estimation results based on real-time sampling data is designed.(3)Design engineering deployment plans for the overall system.Engineering plan design is the landing link of research work and has important significance.This paper analyzes the existing data storage and processing framework,and selects the solutions that are suitable for the deployment of this system to design the system architecture;the cluster adopts the method of mixed deployment to improve the CPU utilization of the machine.In order to make cluster deployment more efficient,a deployment optimization scheme based on service attributes is proposed.In summary,this paper studies GPS trajectory data,and proposes corresponding solutions to problems such as road matching,road transit time estimation,missing data completion,and system deployment scheme optimization.The experiments prove that these schemes have practical value and can be used for roads.Transportation optimization provides effective support.