Dynamic Vehicle Perception Using Automotive Radar

Vehicle perception, namely other vehicle detection and tracking is one of the most critical part in autonomous vehicle system. In recent year, automotive radars have been introduced successfully into this field with sensors in the 24 GHz frequency domain for short range perception and 77 GHz for long range perception. In this paper, a 77 GHz long range millimeter wave radar (LRR) is utilized to detect and track the vehicle simultaneously because we focus on the vehicles moving in front of the ego car at a relatively long distance. In order to approach reality application, an advance constant turn rate and velocity magnitude (CTRV) dynamic model is employed to simulate the target vehicle kinematics. 2D-FFT is used to obtain the velocity and range of the target vehicle. Then, we employ the unscented Kalman filter (UKF) to estimate the state of the detected vehicle since its non-linear motion. The whole presented automotive radar system is validated in simulation environment based on some reality scenarios. The results show that the performance of the framework is promising and feasible.