Mobile Crowdsensing Based Traffic Velocity Missing Data Recovery Algorithm

With the increase of vehicle ownership,traffic congestion has become an urgent problem in modern cities.To reduce the burden of road network,effective traffic assessment is important for traffic management.However,traditional traffic sensors have high cost of deployment and maintenance.Along with the rapid development and upgrading of mobile terminal processor,memory and other hardware devices,real-time traffic conditions can be achieved by combining real-time road data obtained by on-board GPS with urban traffic network information,thus providing better traffic information services for people.It has become an effective way to solve traffic congestion and other problems at present.However,the limited size of the vehicles and their uneven spatiotemporal distribution lead to the lack of data in the use of traffic assessment,which is one of the main problems it faces in the application of urban road network traffic assessment.Therefore,in this thesis,we investigate the recovery of traffic missing velocity data based on mobile crowdsensing.The main contributions can be summarized as follows:(1)The traffic velocity matrix based on the probe vehicle data is extracted.In order to obtain the road network information,we propose a network extraction algorithm based on grid density.Specifically,the pre-processed data is gridded and the road center point is extracted according to its density degree,and then the road network is formed by fitting the center point,and the obtained road network is vectorized to get the road network data.After that,according to the characteristics of GPS data,an adaptive calculation method of road velocity is designed.Based on the average travel velocity,the traffic condition is fully considered.The traffic velocity in the road section is calculated by classification,and the traffic velocity matrix representing traffic conditions is obtained.(2)A traffic velocity data recovery algorithm based on compression sensing is designed.In this thesis,according to the missing location of different data,the missing data can be classified into three categories: random missing,whole missing and rowcolumn missing.On this basis,an improved compressed sensing algorithm is proposed to recover the missing data by designing a sparse basis with the consideration of the spatialtemporal correlation between the data.Then,the prediction problem of traffic velocity missing data is modeled as the recovery problem of sparse vector,and different measurement matrices are selected according to different data missing conditions.Finally,the recovery of traffic missing data is realized.(3)The performance of the proposed algorithm is fully verified based on large-scale real taxi dataset.Four typical data recovery algorithms of SI、KNN、3D-SHAPE、GM(1,N)are selected,and multiple indexes are employed to compare the performance with the proposed STC-CS algorithm in this thesis.The experimental results show that the proposed STC-CS algorithm can recover the missing data accurately when the degree of data missing is more than 50%,and the performance is better than that of other similar algorithms.In the case of whole missing and row-column missing,the proposed STC-CS algorithm can recover the missing data more accurately,and the efficiency of the algorithm outperforms that of other algorithms.It can provide a good solution for traffic condition assessment.