Abstract:

This review article provides an overview of radar sensing for robust, cost-effective perception in industrial and mobile robotics. Radar sensors provide a spatio-dynamic signature of objects via object detection, localization, and velocity estimation, while remaining robust under adverse environmental conditions. Furthermore, edge and on-device processing in radar sensors enables real-time perception and decision-making for robots. We analyze diverse use cases of autonomous robots such as robot arms, autonomous mobile robots (AMRs) and humanoid robots in context of industrial deployment such as warehouse automation and manufacturing. We compare radar with alternative modalities such as LiDARs, cameras, and ultrasonic sensors. We show how new technology developments in imaging radars a,long with advanced machine learning (ML) methods support robotic perception functions such as localization and mapping, 3D collision avoidance, and human detection. We discuss radarspecific challenges such as sparse point cloud data and multipath clutter and how radar-tailored ML models and sensor fusion can address them. We provide a comprehensive overview of recent advances in radar edge processing and identify key open challenges like the need for environment-aware radar processing, cross-modal learning methods, low-power hardware acceleration, availability of public radar datasets for robotics use cases.