Reliable tracking of road users plays a critical part on the way to safe automated driving. In this paper, a machine learning based tracking approach on radar data is presented utilizing the radar target point clouds from multiple time steps as input to detect road users and to predict their tracking information. The detection and tracking of objects is achieved by applying a combination of known feature extractors from lidar and camera detection tasks. The generated feature maps are used as input to two branches - one branch for detection and one for tracking. In experiments on an extensive real-world radar data set, the proposed model achieves promising results in tracking performance compared to a basic clustering and a classification assisted tracking approach.

Automotive radars are becoming increasingly popular sensors for vehicle self-localization tasks because of their robustness and affordability. Newly available polarimetric sensors provide rich information about the occurring scattering mechanisms and thus enhance the uniqueness of landmarks leading to increased reliability. This paper presents a novel approach to incorporate the polarimetric information into a gridmap based on covariance matrices. Thereby, knowledge about up to three different scattering mechanisms per cell can be preserved. Such a discrimination is essential to account for the changing scattering information caused by varying viewing angles with respect to the landmark, which result from the vehicle passing by on the one hand and different mounting positions in a multi-sensor fusion setup on the other. To demonstrate suitability of the covariance representation, an evaluation of the proposed approach based on real-world experiments is presented. A substantial improvement in landmark recognition performance in difficult situations is achieved by taking advantage of the polarimetric covariance information.