CSF: Contextual Safety Filtering for Motion Generators

2026-10-08

AI Summary

This study addresses the lack of scene-aware safety mechanisms in text-driven motion generation by proposing a training-free safety filtering framework. The framework introduces a novel reference-trajectory-based affine safety value formulation that translates natural language rules into Control Barrier Function Quadratic Programming (CBF-QP) constraints, thereby achieving scene-adaptive safety control. Notably, the proposed method can be deployed without requiring additional annotations or modifications to model weights. Experimental results demonstrate that this framework reduces hazardous event rates by 90% across four pretrained generators while preserving 88–100% of benign motions. Furthermore, real-world safety is validated through deployment on a Unitree G1 humanoid robot, confirming the practical effectiveness of the approach in physical scenarios.

Abstract

Text-conditioned motion generators produce trackable whole-body motion, but they have no notion of scene-dependent safety: the same action may target an object or a person. Existing safeguards either inspect the prompt, require labeled motion data, or enforce geometric constraints; therefore, they do not directly account for how scene context changes a motion's meaning. We introduce contextual safety filtering (CSF), a training-free filter that grounds natural-language safety rules in safe and unsafe reference trajectories produced by the generator. For each active rule, safe and unsafe reference trajectories define an affine safety value that a safe reference tracking CBF-QP enforces. Across four pretrained generators with different architectures, CSF activates the intended rules in all explicit and scene-triggered unsafe cases and reduces the danger-event rate by up to 90%, while preserving 88-100% of benign motions. We demonstrate the complete system on a real-world Unitree G1, where it successfully prevents unsafe motions in a variety of scenarios, including interactions with humans and objects.

Problem

• motion generation
• contextual safety
• text-conditioned motion
• safety filtering
• scene context

Innovation

• Contextual Safety Filtering
• Training-free
• Control Barrier Function
• Motion Generation
• Natural Language Safety Rules