This research introduces a semantic safety filter that enables robots to understand and adhere to common-sense safety rules in human environments, preventing mistakes like spilling water on electronics.

• Combines semantic scene understanding with motion safeguards
• Implements real-time safety verification using control barrier functions
• Demonstrates effective safety constraints in multiple scenarios with minimal intervention
• Successfully prevents unsafe actions while allowing goal completion

For Engineering teams, this represents a significant advancement in robot safety systems that can operate alongside humans without requiring exhaustive pre-programmed rules, enabling more flexible deployment in dynamic environments.

Semantically Safe Robot Manipulation: From Semantic Scene Understanding to Motion Safeguards