Research On Method For Traffic State Identification Based On Matrix Completion Theory In Connected Autonomous Vehicle And Highway System

In recent years,with the advancement of China’s economic development and urbanization,the living standards of residents are improving day by day,and the number of cars continues to grow.However,it also leads to increasingly serious problems such as traffic congestion,energy consumption,environmental pollution and noise pollution.The supply of transportation infrastructure is difficult to meet the growing travel needs of residents.At the same time,with the development of new technologies such as communication technology,big data and artificial intelligence,the transportation system has begun to develop in the direction of intelligence and connectivity.Both theory and practice have proved that intelligent and connected transportation system can help to improve traffic congestion,reduce energy consumption and other social problems.Traffic state recognition is the premise of the implementation of traffic management measures.Real time,comprehensive and reliable road traffic state recognition is an important support of active traffic management system.It can also provide travel reference for traffic travelers and improve traffic safety and operation efficiency.In recent years,connected vehicles with various sensors have developed rapidly.The future traffic system will be the mixed state of connected vehicles and manually driven vehicles.The sensors loaded by connected vehicles can effectively perceive the traffic information of surrounding vehicles and provide data basis for road traffic state recognition.Most of the traditional estimation methods can only obtain the traffic state at the section level,it has low accuracy.As a new traffic information detection equipment,connected vehicles have the advantages of wide coverage and low maintenance cost.Therefore,Lane level traffic state estimation can be realized to meet the needs of fine control of intelligent traffic system,Therefore,using connected vehicles’ detection data,this paper studies the traffic state recognition based on matrix completion algorithm in intelligent and connected environment.Firstly,the information collection experiment is carried out,the differences of collection methods of different traffic detection modes are explained,and the reasons for selecting connected vehicles for data collection are explained.Then,the intelligent and connected environment described in this paper is defined.NGSIM datasets are selected as data source for connected vehicles collecting data.Then,aiming at the problem of data noise in the dataset,Kalman filter algorithm is used to filter the NGSIM dataset.Then the correlation coefficient matrix and grey correlation analysis method are used to analyze the time correlation and spatial correlation of traffic flow to pave the way for the following matrix completion method based on time and spatial correlation.Secondly,by setting the sensor distribution mode of connected vehicles,the information collection algorithm of the connected vehicles is designed,and the data collected by the connected vehicles is analyzed in different penetration,different sensor distribution modes and different traffic state conditions,etc.Then,the traffic state recognition method in the intelligent and connected environment is constructed based on matrix completion theory.The spatio-temporal division modeling of road sections is carried out and section mean speed is selected to represent traffic state of each spatialtemporal interval,then three traffic data loss modes are summarized.Improved singular value thresholding algorithm is proposed to estimate the missing information in spatio-temporal intervals,this method is contrasted with singular value thresholding algorithm and improved K nearest neighbor algorithm,and the applicability of these methods is explained.Finally,SVT algorithm,improved SVT algorithm and improved KNN algorithm are empirically analyzed by using the state information perceived by connected vehicles.The results show that the improved SVT algorithm has higher accuracy.To the problem of the estimation accuracy of random missing interval,k-nearest neighbor weighted clustering search algorithm is designed,results show that this algorithm can effectively improve the estimation accuracy of random missing interval.Take the dataset of I80-1600-1615 as an instance,traffic state identification is conducted and results show that proposed algorithm can effectively reshow the traffic state in smaller particle size.