AI Insight
This paper presents X-Safe, a new safety system for robotic manipulation that prevents collisions across different robot types without requiring robot-specific safety models or simplified approximations. The approach works in configuration space and uses action masking to restrict robot policies to safe regions, providing formal probabilistic safety guarantees. Testing on two robot embodiments and three different control policies showed less than half the performance degradation compared to existing safety methods while maintaining collision avoidance guarantees.
Why it matters
This work addresses a major challenge in deploying learning-based robot controllers safely across different robot platforms. The embodiment-agnostic approach could reduce the engineering effort required to deploy safe manipulation policies on new robots, potentially accelerating adoption of learning-based control in industrial and service robotics applications.
Understand the Science
⚠️ Preprint – Noch nicht peer-reviewed
Dieser Artikel wurde noch nicht von unabhängigen Experten begutachtet. Die Ergebnisse sind vorläufig und sollten mit Vorsicht interpretiert werden.
-cross
Abstract: Ensuring safety of learning-enabled robotic manipulation across diverse embodiments and tasks still requires significant manual engineering. For collision avoidance, existing approaches typically rely on embodiment-specific learned safety approximations or simplified robot models together with stopping-based fallback controllers that compromise performance. In this paper, we propose an embodiment-agnostic safeguarding approach, X-Safe, which provides formal probabilistic guarantees for collision avoidance without model simplifications or learned safety functions. By operating in the configuration space, our safeguarding method works with different types of robot policies. X-Safe reasons locally around the current configuration to establish safe regions and uses action masking to restrict the policy’s support to these safe regions. We demonstrate X-Safe on two embodiments and three policies, both in simulation and on hardware. We observe less than half of the performance degradation compared to state-of-the-art safeguarding, and empirically corroborate our formal collision avoidance guarantees.
Source: X-Safe: Embodiment-Agnostic Safeguarding for Robot Arm Manipulation Policies via Action Masking