Learning Fast, Tool-Aware Collision Avoidance for Collaborative Robots
Joonho Lee, Yunho Kim, Seok Joon Kim, Van Quan Nguyen, Young Jin Heo
AI summary
Problem
Current controllers assume fixed tools and full visibility, leading to collisions or overly conservative behavior in dynamic, contact-rich settings with occlusions.
Approach
A learned 3D CNN perception model filters robot/tool components and estimates collision risk under partial observability, driving a constrained reinforcement learning policy that seamlessly switches to classical inverse kinematics when safe.
Key results
- Sub-millisecond reactive avoidance with sub-millimeter tracking accuracy
- ~60% lower computational cost than state-of-the-art GPU planners
- Real-time adaptation to varying tool geometries and contact modes
- Robust obstacle avoidance under partial observability in simulation and real-world tests
Why it matters
Enables reliable, high-speed deployment of collaborative robots in dynamic human-centric environments where traditional planners fail.
Abstract
Ensuring safe and efficient operation of collabo- rative robots in human environments is challenging, especially in dynamic settings where both obstacle motion and tasks change over time. Current robot controllers typically assume full visibility and fixed tools, which can lead to collisions or overly conservative behavior. In our work, we introduce a tool-aware collision avoidance system that adjusts in real time to different tool sizes and modes of tool-environment interaction. Using a learned perception model, our system filters out robot and tool components from the point cloud, reasons about occluded area, and predicts collision under partial observability. We then use a control policy trained via constrained reinforcement learning to produce smooth avoidance maneuvers in under 10 milliseconds. In simulated and real-world tests, our approach outperforms traditional approaches (APF, MPPI) in dynamic environments, while maintaining sub-millimeter accuracy. Moreover, our system operates with approximately 60 % lower computational cost compared to a state-of-the-art GPU-based planner. Our approach provides modular, efficient, and effective collision avoidance for robots operating in dynamic environments. We integrate our method into a collaborative robot application and demonstrate its practical use for safe and responsive operation.