Adaptive Beamforming for Situation-Aware Automotive Radars Under Uncertain Side Information

Abstract

Radar is an important sensing modality that supports advanced levels of assisted and autonomous driving. In this work, we exploit side information, such as lane topology maps of the environment, position, and orientation information of the ego vehicle, to design beamformers in automotive radars. Specifically, we present a convex optimization-based method for transmit beamformer design using location-based static environment maps derived from georeferenced maps. The designed beams allocate less power along the directions where a static obstacle in the environment is closer and vice versa. We study the robustness of our situation-aware transmit beamforming technique to uncertainties in the position and orientation information of the ego vehicle. We also address these uncertainties by extending our situation-aware beamforming approach using tools from stochastic optimization (SO). Through simulations on the public dataset nuScenes, we show that our method achieves better detection than situation-agnostic radar sensing. Furthermore, our design is robust against errors in estimating the position and the orientation of the ego vehicle.