Research On Signal Processing Algorithm Of Millimeter Wave Radar Based On FPGA

Nowadays,Frequency-Modulated Continuous Wave(FMCW)millimeter-wave radar is being increasingly used in the field of autonomous driving.On the one hand,target detection is being an important part of FMCW radar.On the other hand,the interference caused by the FMCW radars has become more and more serious,which means the detection of interference is becoming much more important.In summery,the algorithm of Constant False Alarm Rate(CFAR)and change-point detection in FMCW radar signal processing has become the focus.For target detection,the CFAR detection algorithm has excellent detection performance,but the problem is that a single CFAR detector cannot cope with the complex millimeter wave radar detection environment.For interference detection,change-point detection has accurate detection performance.However,due to the serial operation and the high complexity of change-point detection,it is difficult to achieve real-time detection on millimeter-wave radars,which cannot meet the high-performance real-time processing of millimeter-wave radars.Field programmable gate arrays(FPGAs)have reconfigurable features and design flexibility.FPGAs also have the advantage of being able to accelerate signal processing algorithms in parallel.Therefore,the problems of CFAR detection and change-point detection can be solved by using the reconfigurable and parallel features based on FPGA.The main work of this thesis is as follows:(1)The CFAR detection and change point detection algorithms are analyzed.For the CFAR detection algorithm,the principles of CA-CFAR and OS-CFAR algorithms are studied,and the strengths and weaknesses of CA-CFAR and OS-CFAR are discussed.These CFAR algorithms are simulated for target detection.In addition,the problem of target "masking effect" generated by CA-CFAR is analyzed.For change point detection,the basic concept of the change point detection is studied,and the principles of the three change point detection algorithms of sliding window(Win),binary segmentation(Bin Seg)and bottom-up segmentation(BOTUP)are analyzed.The problem of insufficient real-time performance in change-point detection is analyzed,and verify the detection effect through simulation experiments.(2)A reconfigurable CFAR structure is proposed.A reconfigurable CFAR structure(CAOS-CFAR)is designed based on FPGA: different CFAR algorithm is used according to different clutter detection environment of FMCW radar.The CA-CFAR detector is selected in a uniform clutter environment,and the OS-CFAR detector is selected in a non-uniform environment.The FPGA-based reconfigurable CFAR detector can not only effectively maintain the target detection performance of millimeter-wave radar,but also can greatly reduce resource consumption of FMCW system.(3)An accelerated structure of BOTUP(ACC-BOTUP)is proposed.The ACC-BOTUP is realized based on FPGA.In ACC-BOUTP,since BOTUP has sub-functions realized by multiple cost functions with similar structures,a parallel structure is proposed to run the cost functions in parallel,which greatly reduces system latency.In addition,a deep pipeline structure is proposed to further reduce the latency.Compared with the original architecture of BOTUP,ACC-BOTUP reduces latency of the system by 82%,which can improve the performance of real-time detection.