Taming VLAs under Robot Execution Errors: Self-Compensation and Stress Testing

2026-09-29

AI Summary

This study addresses task failures in Vision-Language-Action (VLA) models caused by robotic execution errors by proposing a self-compensating VLA framework. The method employs an online update strategy based on execution residuals to pre-compensate for errors, enabling deployment-time adaptation without requiring reward labels. Additionally, we introduce RoboStress, a simulation benchmark encompassing diverse mechanical faults to systematically evaluate model robustness. Experimental results demonstrate that the proposed framework improves task success rates by over 30 percentage points in both simulated and real-world bimanual scenarios, while exhibiting strong generalization capabilities to unseen objects.

Abstract

Vision-language-action (VLA) policies often fail when a robot's executed motion deviates from their commanded action. Such execution errors arise from the robot's mechanics and operating conditions, such as wear and payload changes. We propose self-compensating VLA, a deployment-time adaptation method that enables a VLA policy to pre-compensate for the robot's execution errors when generating commands. Without task rewards or labels, it updates the policy online using the residual between the action commanded by a VLA and the motion executed by the robot. To stress-test VLA robustness across execution conditions that are impractical to cover with physical robots alone, we introduce RoboStress, a controlled simulation benchmark. It combines established joint-level models of friction, backlash, compliance, and gravity-compensation error into seven deployment scenarios whose execution errors depend on the robot's state and motion history. On RoboStress, self-compensating VLA achieves higher average task success than both the base policies and methods that build in robustness during training. On two physical robot arms with different usage histories, it raises the average task success rate by more than 30 percentage points on each arm, and the gains extend to objects not seen in the task demonstrations.

Problem

• Vision-language-action (VLA)
• execution errors
• robot robustness
• stress testing

Innovation

• Vision-Language-Action (VLA)
• Self-Compensation
• Execution Errors
• RoboStress Benchmark
• Online Adaptation