Data Association and Tracking for Automotive Radar Networks

Radar sensors in the 24- and 77-GHz frequency domain will be used to increase comfort and safety in many future automotive applications. In this paper, a radar network with four short-range radars is considered. Each sensor measures individually only the range information of all targets inside the observation area. The Cartesian coordinates of each target are calculated by a trilateration technique based on range measurements selected in a data-association procedure. Estimating a target position based on range measurements is called trilateration. In contrast to this, estimation of a target position based on pure angular measurements is called triangulation. In automotive applications, situations with multiple targets almost always occur. Therefore, a high-performance data association is very important to separate and to distinguish between these targets. To avoid errors in the data-association step and resulting ghost targets, this paper describes a technique that combines the procedures of data association and position estimation into a single step. This signal-processing technique shows very good results in multitarget situations and reduces the number of ghost targets drastically.