Cartesian Tracking for Advanced Driver Assistance Imaging Radar Systems

Abstract

This paper presents an approach of using the Kalman Filter to track automotive radar detections in Cartesian space, useful for Radar-Vision fusion. A typical automotive radar would extract, for every detection, its respective range, radial velocity and Direction of Arrival. The problem at hand is that having no information on the Direction of Displacement, one cannot uniquely map the measurements to cartesian coordinates, which is required for the prediction step of the Kalman Filter. The proposed method uses Recursive Error Division filters and a stability-controlled feedback loop to iteratively estimate the DoD, making Cartesian tracking possible.