Robot Collision Avoidance

Robot collision avoidance studies how robots can move and act without hitting obstacles, people, or other robots, including in cluttered, changing environments. Methods range from geometric planning and model predictive control to control barrier functions that filter unsafe commands, alongside learned policies using depth, vision, tactile sensing, or motion forecasts. Work spans drone flight, manipulation, legged robots, human-robot collaboration, and multi-robot coordination, with evaluation increasingly combining simulation, benchmarks, and physical tests. A central tension is maintaining formal or empirical safety under uncertain sensing and dynamics without making behavior too conservative, slow, or likely to stall.