PhysReflect-VLA: Physical Feasibility and Self-Reflective Regulation for Reliable Vision-Language-Action Policies

2026-06-25

AI Summary

Long-horizon robotic manipulation often fails due to physically infeasible actions, contact-induced disturbances, and the absence of online self-correction mechanisms. This work proposes a plug-and-play runtime reliability framework that, for the first time, integrates physical consistency verification and a large language model (LLM)-driven online self-reflection mechanism into the closed-loop control pipeline of vision-language-action (VLA) policies. The framework employs a feasibility operator, an action interpretation operator, and an LLM-based reflection module, trained via a two-stage strategy, to enable dynamic error diagnosis and correction. Evaluated on multi-stage, high-contact real-world tasks, the approach improves average task success rates by 5.4% over baseline VLA models, significantly enhancing stage-wise stability and overall robustness.

Abstract

Long-horizon robotic manipulation is highly sensitive to physically infeasible transitions, contact-induced disturbances, and the lack of effective self-correction during execution. Although Vision-Language-Action (VLA) models provide strong task grounding through multimodal learning, they typically generate actions in a feed-forward manner without explicitly checking physical feasibility or diagnosing execution errors online. We present PhysReflect-VLA, a plug-and-play execution-time reliability framework that augments VLA policies with physical feasibility evaluation and structured self-reflection in a closed-loop control pipeline. A Feasibility Operator evaluates whether candidate actions induce dynamically consistent state transitions; an Action Explanation Operator verifies transition coherence; and an LLM-based Reflection Module analyzes state discrepancies to generate corrective guidance for subsequent actions. A two-stage training procedure stabilizes feasibility modeling and integrates reflection into the control loop. Experiments on multi-stage, contact-rich real-world manipulation tasks show consistent improvements in stage-wise stability and overall task success compared with representative VLA baselines with an average gain of 5.4\%. Ablation results further indicate that feasibility checking and reflection-based correction both contribute to improved execution robustness. These results highlight the importance of embedding physical consistency checks and online self-reflection for reliable long-horizon robotic manipulation.

Problem

• long-horizon robotic manipulation
• physical feasibility
• self-correction
• execution errors
• contact-induced disturbances

Innovation

• physical feasibility
• self-reflection
• vision-language-action
• closed-loop control
• robotic manipulation