Detection and Backprojection of Ghost Targets within a Network of Radar Sensors

Trajectory planning for autonomous vehicles is based on reliable and accurate mapping. Robust and weather-independent mapping is possible, for example, with radar sensors. In this context, target localization is based on the assumption that electromagnetic waves are emitted by the radar sensor, reflected at the target, and received at the radar sensor. However, as soon as a reflecting wall such as a guard rail is located at the roadside, ghost targets can occur. The presented algorithms allow not only ghost target detection but also ghost target suppression. Furthermore, a method is presented which allows to use ghost targets to improve the localization of real targets or to detect hidden real targets. Since the position of the wall has to be known for this purpose, an algorithm for accurate pose estimation of the wall with localization errors smaller than 7 cm and orientation errors smaller than 0.2° is presented. Finally, all three algorithms are evaluated by means of a measurement.