Research Analyzer
← Back ICRA 2026

SEEC: Stable End-Effector Control with Model-Enhanced Residual Learning for Humanoid Loco-Manipulation

Jaehwi Jang, Zhuoheng Wang, Ziyi Zhou, Feiyang Wu, Ye Zhao

PDF

AI summary

Key figure (auto-extracted from paper)
A model-enhanced residual learning framework stabilizes a humanoid robot's arm during dynamic walking by compensating for base-induced disturbances, enabling robust control across unseen gaits and real-world deployment.
Humanoid control loco-manipulation reinforcement learning end-effector stabilization sim-to-real residual learning

Problem

Stabilizing a humanoid robot's arm end-effector during dynamic locomotion is difficult due to high degrees of freedom and base-induced disturbances. Existing controllers either rely on inaccurate dynamics models or overfit to training conditions, failing to generalize to new walking patterns.

Approach

SEEC decouples upper and lower-body control, training a reinforcement learning policy to compensate for locomotion disturbances using model-derived acceleration signals and a synthetic perturbation generator. This allows the arm controller to learn robust stabilization independent of the specific walking gait.

Key results

  • Up to 60% reduction in end-effector acceleration compared to IK and RL baselines
  • Robust generalization to unseen locomotion controllers without retraining
  • Successful zero-shot sim-to-real deployment on the Booster T1 humanoid
  • Stable performance in diverse loco-manipulation tasks like chain-holding and object carrying

Why it matters

Provides a modular, robust control strategy that enables humanoids to perform precise manipulation while walking, accelerating practical deployment in dynamic environments.

Abstract

Arm end-effector stabilization is essential for hu- manoid loco-manipulation tasks, yet it remains challenging due to the high degrees of freedom and inherent dynamic instability of bipedal robot structures. Previous model-based controllers achieve precise end-effector control but rely on precise dynamics modeling and estimation, which often struggle to capture real-world factors (e.g., friction and backlash) and thus degrade in practice. On the other hand, learning-based methods can better mitigate these factors via exploration and domain randomization, and have shown potential in real-world use. However, they often overfit to training conditions, requiring retraining with the entire body, and still struggle to adapt to unseen scenarios. To address these challenges, we propose a novel stable end-effector control (SEEC) framework with model-enhanced residual learning that learns to achieve precise and robust end-effector compensation for lower-body induced disturbances through model-guided reinforcement learning (RL) with a perturbation generator. This design allows the upper-body policy to achieve accurate end-effector stabilization as well as adapt to unseen locomotion controllers with no additional training. We validate our framework in different simulators and transfer trained policies to the Booster T1 humanoid robot. Experiments demonstrate that our method consistently outperforms baselines and robustly handles diverse and demanding loco-manipulation tasks. More details and videos are available at: https://seec-humanoid.github.io. 2026 IEEE International Conference on Robotics and Automation (ICRA 2026) June 1-5, 2026. Vienna, Austria 979-8-3315-8160-2/26/$31.00 ©2026 IEEE 1989

Index terms

Legged Robots Reinforcement Learning Mobile Manipulation

Related papers