Probabilistic SAR Processing for High-Resolution Mapping Using Millimeter-Wave Radar Sensors

In the field of autonomous driving, highly accurate representations of the environment are essential for trajectory planning as well as for estimating the vehicle's location. Today, this can be achieved with the help of chirp-sequence radar sensors or radar sensor networks. The possibilities for environmental mapping cover simple point clouds, target-list-based grid maps, and raw data-based high-resolution synthetic aperture radar (SAR) maps. While for target-list-based grid maps, it has already been shown that a probabilistic occupancy grid map has significant advantages over an amplitude-based grid map in terms of robustness and resolution, and no probabilistic approaches to SAR processing exist up to now. This article presents a fundamental approach of processing radar raw data to generate high-resolution SAR images based on probabilities. A probabilistic SAR processing is presented that combines high-resolution environmental mapping with amplitude-independent target detection. Based on the measurements, a qualitative and quantitative comparison between conventional amplitude and phase-based SAR processing, the presented probabilistic SAR processing, and a probabilistic target-list-based occupancy grid map is performed. Since the presented algorithm is not limited to the automotive field and chirp-sequence radar sensors, it can be extended to arbitrary SAR applications and radar architectures.