ActSafeGuard keeps robot actions safe without lowering success
ActSafeGuard: Differentiable and Training-Aligned Constraint Enforcement for Flow-Matching Policies
RoboticsArtificial Intelligence
Summary
Robots that interact with the real world need to make sure their actions don’t break physical limits, or else they risk causing harm or failing tasks. The paper’s authors created ActSafeGuard, a new method that helps robot action models learn to stay within these safety limits during training, not just at use time. This approach uses math to smoothly guide the robot’s decisions, ensuring all actions are safe while still succeeding at their goals. Tests showed the method reliably produced safe actions and sometimes even improved task success.
What this means in practice
- •For robotics engineers: Ensure robot control policies never command unsafe actions while maintaining or improving task performance.
- •For industrial automation teams: Deploy robots that strictly respect physical safety constraints on the factory floor without complex safety overrides.
Authors
Jianming Ma, Rongjun Jin, Xiaxi Si, Yang Zhang, Yiheng Li, Yue Gao
Abstract
Vision-Language-Action (VLA) and World-Action Models (WAMs) have demonstrated strong capabilities in general-purpose robotic manipulation, yet their generated actions may violate hard physical constraints and therefore be unsafe or infeasible for deployment. Existing safety approaches either optimize statistical safety objectives without deterministic per-step guarantees or correct unsafe actions only during inference, creating a mismatch between policy training and execution. We introduce ActSafeGuard, a differentiable and training-aligned safeguard layer for flow-matching based policies. ActSafeGuard integrates hard action feasibility into policy learning, not merely treating safety as an inference-time external component. Through an analytical ray-scaling operator design, ActSafeGuard enables boundary-aware gradients to guide the model to naturally learn constrained manifolds. Extensive experiments on multiple standard foundation backbones ($π_{0.5}$ and Fast-WAM) across various tasks demonstrate that ActSafeGuard consistently achieves a $100\%$ step safety rate while fully preserving or even boosting task success rates, providing a scalable and minimally invasive solution for safe embodied AI deployment.