OmniTacTune: Policy-Agnostic Real-World RL for Tactile Residual Adaptation of Visual Policies

2026-07-04

AI Summary

This work addresses the limitations of vision-based policies in contact-rich tasks, which often fail due to the absence of local force and contact geometry information, while tactile sensing suffers from high data acquisition costs and poor generalization. The authors propose a policy-agnostic real-world reinforcement learning framework that integrates tactile feedback into a pretrained visual policy through a lightweight tactile residual correction mechanism. The approach employs a two-stage training process: first, a base visual policy autonomously collects data to guide tactile representation learning; then, a residual policy is optimized via online interaction. This method achieves, for the first time, cross-task, cross-policy, and cross-tactile-representation adaptation of tactile residuals without retraining the main policy. In four real-world contact-intensive tasks, it boosts success rates from 5–40% to 85–100% with only 40–80 minutes of online interaction.

Abstract

Visual policies learned from human videos, teleoperation, and robot demonstrations offer scalable motion priors, but often fail in contact-rich manipulation, where success significantly depends on local force and contact geometry. Tactile sensing provides these complementary signals, yet tactile data remain costly to collect and hard to generalize across sensors, robots, and tasks. We introduce OmniTacTune, a policy-agnostic real-world RL pipeline that adapts tactile feedback to pretrained visual policies through residual correction. OmniTacTune uses a two-stage design: it first bootstraps tactile-aware learning from autonomous base-policy rollouts, then learns a lightweight tactile residual policy through online interaction. Extensive experiments show that OmniTacTune generalizes across diverse contact-rich tasks, visual base policies, and tactile representations. Across four real-world contact-rich tasks, it improves visual base policies from 5-40% success to 85-100% within 40-80 minutes, demonstrating an efficient path for adapting tactile feedback to scalable visual robot policies. Project page: https://colinyu1.github.io/omnitactune-site/

Problem

• tactile sensing
• visual policies
• contact-rich manipulation
• real-world reinforcement learning
• policy adaptation

Innovation

• tactile residual adaptation
• policy-agnostic RL
• visual-tactile fusion
• real-world reinforcement learning
• contact-rich manipulation