VLSA: Vision-Language-Action Models with Plug-and-Play Safety Constraint Layer

TL;DR

This paper introduces VLSA, a safety-enhanced vision-language-action model with a plug-and-play safety layer, demonstrating improved obstacle avoidance and task success in robotic manipulation tasks within safety-critical environments.

Contribution

The work presents a novel safety constraint layer integrated into VLA models, providing theoretical safety guarantees without compromising instruction-following capabilities.

Findings

59.16% improvement in obstacle avoidance

17.25% increase in task success rate

Outperforms state-of-the-art baselines in safety-critical tasks

Abstract

Vision-Language-Action (VLA) models have demonstrated remarkable capabilities in generalizing across diverse robotic manipulation tasks. However, deploying these models in unstructured environments remains challenging due to the critical need for simultaneous task compliance and safety assurance, particularly in preventing potential collisions during physical interactions. In this work, we introduce a Vision-Language-Safe Action (VLSA) architecture, named AEGIS, which contains a plug-and-play safety constraint (SC) layer formulated via control barrier functions. AEGIS integrates directly with existing VLA models to improve safety with theoretical guarantees, while maintaining their original instruction-following performance. To evaluate the efficacy of our architecture, we construct a comprehensive safety-critical benchmark SafeLIBERO, spanning distinct manipulation scenarios characterized by varying degrees of spatial complexity and obstacle intervention. Extensive experiments demonstrate the superiority of our method over state-of-the-art baselines. Notably, AEGIS achieves a 59.16% improvement in obstacle avoidance rate while substantially increasing the task execution success rate by 17.25%. To facilitate reproducibility and future research, we make our code, models, and the benchmark datasets publicly available at https://vlsa-aegis.github.io/.