Any-Body Guard: Universal Safeguarding for Manipulation Policies via Action Masking

2026-06-20

AI Summary

Existing safety assurance methods for robotic manipulation often prove overly conservative, computationally expensive, or reliant on extensive engineering adaptation when applied across diverse robot embodiments and tasks, thereby struggling to balance safety and performance. This work proposes X-Safe, a framework that integrates object-level quasi-static scene representations with forward kinematics in configuration space to enable general-purpose, real-time collision avoidance through action masking. X-Safe provides the first formal probabilistic safety guarantee without requiring additional data or embodiment-specific engineering, eliminating dependence on complex controllers or fine-grained scene models. Experimental results demonstrate its effectiveness across multiple robots and policies, achieving zero collisions in hardware trials, significantly smaller performance degradation compared to existing approaches, and empirical validation of its theoretical safety guarantees.

Abstract

Ensuring safety of learning-enabled robotic manipulation across diverse embodiments and tasks still requires significant manual engineering. Existing approaches typically rely on heuristically designed fallback controllers or complex forward invariance assessments. These methods are often too conservative for task success, too computationally expensive for real-time execution, too heuristic to provide useful safety guarantees, or too engineering-heavy to transfer between setups. In this paper, we propose a universal safeguarding approach, X-Safe, which reasons directly in the robot's configuration space to provide formal probabilistic guarantees for collision avoidance. By operating in the configuration space, our method transfers across embodiments while relying solely on an object-based, quasi-static scene representation and a forward kinematics model of the robotic manipulator. Thus, X-Safe provides useful formal safety guarantees without requiring additional data, or engineering effort for different embodiments or scenes. We demonstrate X-Safe for diverse embodiments and policies, both in simulation and on hardware. We observe less degradation in task performance compared to state-of-the-art safeguarding, no collisions on hardware experiments, and empirically corroborate our formal guarantees.

Problem

• robotic manipulation
• safety guarantees
• collision avoidance
• universal safeguarding
• configuration space

Innovation

• action masking
• configuration space
• probabilistic safety guarantees
• universal safeguarding
• robotic manipulation