Achieving Cognitive Safety for Human-Robot Collaboration

Human-Robot Collaboration introduces safety challenges that traditional 2D sensors or performance-degrading mitigations cannot adequately solve. This paper introduces the concept of CogniSafe3D, a cognitive safety system that utilizes an external 3D LiDAR to monitor the operational environment. The system's methodology is twofold: it first establishes a deterministic foundation for safety by processing LiDAR point clouds with a Signed Distance Field representation to detect safety zone violations. This core functionality is augmented by a cognitive layer that uses Human Pose Estimation and predictive algorithms to anticipate human movements and infer intent. A Risk Monitoring module synthesizes this information to provide real-time, dynamic risk estimates. This enables the robot to proactively mitigate hazards while reducing unnecessary stops, advancing perception-driven safety in compliance with ISO/TS 15066 standards. The methods developed in the CogniSafe3D research project will be implemented in a TRL-6 prototype.

Keywords: Human-Robot Collaboration, Cognitive Safety, Risk Monitoring, Safety Assessment, 3D LiDAR.