Speed and yaw rate estimation in autonomous vehicles using radar measurements

Abstract: One of the key elements for model-based autonomous driving is the ability of the vehicle to accurately predict its own motion. Knowledge about the motion can then be widely used, e.g., for localization, planning and control. This thesis presents an algorithm that estimates the velocity and the yaw rate based on Doppler radar measurements. This system uses an unscented Kalman filter to extract the motion of the vehicle from multiple Doppler radar sensors mounted on the vehicle. The estimation of these quantities is showed to be critically dependent on outlier detection and the vehicle’s center of rotation. This work presents a framework for detecting dynamical objects, as well as effectively estimating the center of rotation of the vehicle. The proposed system shows, in tests, better root-mean squared error performance than the current employed system by 28.8% and 22.4% for velocity and yaw rate, respectively.

Examiner: Cristian Rojas.