TMA

Overview

TMA adapts a humanoid’s multi-finger manipulation online from tactile interaction history. Trained entirely in simulation, it opens real round doorknobs zero-shot.

Real-World Experiments

Four off-the-shelf round doorknobs, like those in everyday homes, with no real-world fine-tuning.

All videos are fully autonomous and play at 1× speed. They have audio: unmute a clip to hear the spring bounce when it happens.

Select a doorknob to jump to its rollouts

Recovery under disturbance

Recovering and completing the task after slips and external disturbances.

Closer view on knob rotation

Hand-level views of the multi-finger grasp and the turn that releases the latch.

Baselines

10 real-world trials per knob for each method. The reactive policy uses binary fingertip contact without adaptation; RMA adapts from proprioceptive history without touch.

Abstract

Recent humanoid loco-manipulation in articulated environments has largely focused on geometry-dominant interactions, where success is primarily determined by geometric alignment and motion execution. In contrast, forceful multi-finger manipulation requires the robot to adapt to real-world uncertainties in contact and interaction dynamics. We present Tactile-Driven Motor Adaptation (TMA) and an integrated whole-body system for zero-shot sim-to-real forceful humanoid manipulation, and study round-doorknob opening as a challenging benchmark requiring stable multi-finger frictional grasping, sufficient torque generation without slip, force-regulated pushing, and whole-body stability. TMA structures policy learning around interactions observable through tactile feedback, uses cone geometry as an inductive bias for multi-finger contact representation, and adapts motor behavior online from tactile history. An upper-body manipulation policy is trained alongside a pretrained general locomotion controller to coordinate dexterous interaction and whole-body stabilization. Trained entirely in simulation, TMA transfers zero-shot to a Unitree G1 humanoid with BrainCo Revo2 multi-fingered hands and substantially outperforms standard Rapid Motor Adaptation (RMA) and reactive tactile policy baselines across four commercial round doorknobs in the real world.

Supplementary Video

BibTeX

@misc{chen2026tma,
  title  = {{TMA}: Sim-to-Real Forceful Humanoid Manipulation via Tactile-Driven Motor Adaptation},
  author = {Chen, Yiting and Yu, Kelin and Kaingade, Sarthak and Trinh, Bryan and Hang, Kaiyu and
            Mart{\'i}n-Mart{\'i}n, Roberto and Fazeli, Nima and Wu, X. Alice},
  year   = {2026},
  url    = {https://humanoid-tma.github.io/}
}