PEARS: Physical-Prior-Guided Efficient Adaptation via Failure Reasoning and Diffusion Steering for Tactile Manipulation

2026-10-06

AI Summary

This study addresses the performance degradation of pretrained robotic policies in out-of-distribution scenarios and the low sample efficiency of reinforcement learning (RL) post-training by proposing a physics-prior-guided hybrid RL framework. Methodologically, it integrates vision-language models for failure reasoning to dynamically update force constraints, and designs a physics-guided force reasoning module coupled with a high-frequency hybrid force-position controller using tactile feedback. Furthermore, diffusion steering modulates the noise of a frozen base policy to enable online adaptation without parameter updates. Experimental results demonstrate that the proposed approach improves simulation success rates by 12.4%–37.4% while reducing interaction costs by 53.2%. In real-world settings, the method achieves success rates of 95% and 90% on wiping and pipetting tasks, respectively.

Abstract

Pretrained robotic policies can suffer substantial performance degradation under out-of-distribution (OOD) conditions encountered during deployment, motivating post-training through real-world interaction. However, reinforcement-learning (RL)-based post-training typically requires substantial environment interactions, a burden that is especially significant in manipulation, where each trial can be slow, costly, or destructive. Therefore, we present PEARS, a physics-prior-guided hybrid RL framework for sample-efficient online adaptation of pretrained policies with tactile feedback. After each episode, its physics-guided force reasoning (PFR) module uses physical priors encoded in a vision-language model (VLM) to diagnose failures from the visual outcome and tactile interaction history and update task-appropriate contact-force bounds. A high-frequency hybrid force-position controller then enforces these bounds during contact. Complementarily, tactile-conditioned diffusion steering reinforcement learning adjusts the latent noise of the frozen flow-matching policy to correct errors in free-space motion and contact timing without updating the base model. In simulation, PEARS improves success rates by 12.4-37.4 percentage points over the strongest per-task baselines. PEARS also reduces the number of interaction episodes required for a certain success threshold by up to 53.2% relative to the fastest baseline. In real-world experiments, PEARS achieves success rates of 95% on Whiteboard Erasing and 90% on Pipette Liquid Aspiration. These results show that combining the PFR module with policy steering can accelerate adaptation while reducing costly interactions. The project website is available at https://song-kun.github.io/pears.

Problem

• tactile manipulation
• sample-efficient adaptation
• out-of-distribution
• reinforcement learning
• post-training

Innovation

• Tactile Manipulation
• Diffusion Steering
• Failure Reasoning
• Vision-Language Model
• Sample-Efficient Adaptation