Pedestrian Detection and Trajectory-Tracking based FMCW radar For Autonomous Driving

The perception of pedestrian movement is an arduous task for onboard sensors, yet a crucial aspect of autonomous driving development. On this account, this paper presents a novel FMCW 4D radar-based detection and tracking system for pedestrian targets in low signal-to-noise ratio (SNR) conditions. Specifically, the keystone transform-constant false alarm ratio (KT-CFAR) is proposed to correct range migration (RM), and achieve superior target detection. Then we employ the multiple signal classification (MUSIC) algorithm to enhance angle resolution. Lastly, the particle filtering (PF) algorithm is utilized to track pedestrian targets. Through extensive simulation and experimental validation, the proposed method achieves better results, exhibiting outstanding performance with an RMSE of 0.6173 m.