Research On The Optimization Of Interrupted Flow Travel Speed Extraction Based On Dense Signaling Data

Travel speed is an important index of interrupted flow(arterial road and subsidiary road)traffic monitoring and analysis.Traditional loop detection data,video data and floating vehicle data are limited by high installation,high maintenance costs and insufficient coverage.The2/3G signaling data is widely used to extract the travel speed of continuous flow due to its wide spatio-temporal range and low acquisition cost.In these applications,the accuracy is mainly between 80% and 90%.However,due to the limitation of positioning frequency and positioning accuracy,it is seldom used in the interrupted flow(arterial road and subsidiary road),and the accuracy has not reached a consensus.With the development of technology,the positioning frequency of 4/5G dense signaling data has increased significantly,making it possible to apply to interrupted flow(arterial road and subsidiary road),but it also complicates the connection rule between users and base stations,increasing the uncertainty of the extraction accuracy of travel speed under different traffic conditions.To optimize the extraction method of travel speed in interrupted flow,and answer the accuracy problem of interrupted flow travel speed extraction based on signaling data,this paper first constructed an extraction method(data fusion method)of travel speed based on the characteristics of signaling data in a dense environment.The method is mainly composed of a road matching algorithm,signaling data trigger location reconstruction algorithm,and travel speed extraction algorithm.Considering that the traditional projection method and calibration method are difficult to adapt to the large-scale construction of communication base stations in the future,this paper proposes a signaling data trigger location reconstruction algorithm based on a Hidden Markov Chain.In order to extract travel speed more accurately,this paper further considers the delay time of signal lights and constructs an algorithm to extract travel speed between breaks.At the same time,input and output samples of each algorithm are introduced using volunteer data.Thirdly,considering that some specific congestion states are not available sometimes,for the analysis of extraction effect in different traffic conditions,and comparing different methods,this paper construct “traffic – simulation” platform,to generate signaling data of different traffic conditions(unblocked,general congestion and severe congestion state).At the same time,this paper also verifies the validity of the simulation data.The results show that the travel speed extraction accuracy of the proposed method is 81.92%,and the accuracy standard deviation is 6.07%.Compared with the direct method,when only optimize the signaling data trigger points reconstruction algorithm,the accuracy is improved by 14.28 percentage points and the accuracy standard deviation is reduced by 0.77 percentage points,when the signaling data trigger position point reconstruction algorithm and travel speed extraction algorithm are optimized simultaneously,the accuracy is improved by 27.64 percentage points and the accuracy standard deviation is reduced by 3.15 percentage points.Compared with the method based on modified Hellinga,the accuracy is improved by 6.92 percentage points and the accuracy standard deviation is reduced by 1.83 percentage points.The comparison results show that the optimized travel speed extraction method has better performance in dense signaling data environment.