Ego-Motion Estimation Using Radars

Abstract

Vehicle ego-motion can be described by three ego-pose states, three translational ego-velocities and three angular ego-velocities. All nine states can be estimated using range and Doppler velocity measurements from six radars mounted on the ego-vehicle. This thesis proposes a Kalman Filter with a linear measurement model for estimating ego-velocity from Doppler measurements and an Extended Kalman Filter with a nonlinear measurement model for estimating ego-pose from range measurements. Ego-motion groundtruth and radar measurements were simulated using IPG TruckMaker for a set of simple road variations and driving maneuvers. The ego-velocity estimate error is sensitive to how the radars are mounted, for which a design procedure is proposed.