Study On Automotive Radar System For Target Parameter Estimation And Tracking

Today,with the rapid development of advanced assisted and intelligent driving,automotive millimeter-wave radar has become one of the most promising sensors by virtue of its all-day and all-weather charateristic,attracting great attention for domestic and foreign automobile manufacturers and research institutions.Although radar signal processing technology has been relatively mature in the field of military industry,due to different scenarios and cost limitations,developing an automotive millimeter-wave radar with high performance and reliability is still facing great challenges.In this thesis,target parameter estimation and target tracking in automotive millimeter-wave radar are deeply studied based on actual demands and engineering,which have important theoretical significance and practical value..Firstly,based on the LFMCW signal model,the techniques of range,speed and angle measurements are studied successively.For the angle measurement on uniform linear array,the interferometic phase technique,digital beamforming and Capon algorithm are discussed respectively.And the spatial smoothing technique is used to improve Capon algorithm to relax the burden of limited channel data during the automotive radar.For the velocity measurement problem,especially velocity ambiguity,a method of velocity ambiguity resolution based on phase interference is proposed.Secondly,the theories of target tracking,including tracking gate,data association and tracking filtering are discussed.Data association,as one of the key parts in target tracking,is an important and difficult problem.The nearest-neighbor data association algorithm and probabilistic data association algorithm are analysed in this thesis.Based on the characteristics of actual measurements,a new method called Doppler constraint nearest-neighbor data association algorithm is proposed from the perspective of reducing measurements into the gate,which effectively reduces the association uncertainty problem.For tracking,CV model and CA model,which are suitable for weak maneuvering target modeling,are introduced,and then Kalman filtering algorithm and α-β(α-β-γ)filtering algorithm,commonly used in engineering,are studied.Finally,for the BSD radar application,a complete set of multi-target tracking process is designed in this thesis,and some parts and details of tracking are discussed,including a greedy logic track-initiation method using a prior and a prioritized strategy of data association,initialization of targets’ parameter is also studied.In view of the application scenarios,measurement and target classification and scene recognition measures are also proposed.The validity of the multi-tracking algorithm is analyzed and verified using real BSD radar data.