ADAPT: Analytical Disturbance-Aware Policy Training for Humanoid Locomotion

2026-06-15

AI Summary

This work addresses the challenge of degraded motion accuracy and stability in humanoid robots operating in human environments due to external disturbances. The authors propose the ADAPT framework, which employs a physics-based analytical whole-body disturbance observer to estimate residual external forces and torques online without requiring force/torque sensors. These estimates are explicitly fed into a reinforcement learning policy network, endowing the policy with the ability to perceive and adapt to unknown disturbances. By integrating proprioception, robot dynamics modeling, and sensorless external force estimation, ADAPT significantly enhances cross-scenario generalization and robustness. Experiments on the Unitree G1 platform demonstrate that ADAPT achieves superior stability under torso pushes, standing perturbations, and asymmetric hand loads, improves velocity tracking accuracy, and enables light-footed gaits through disturbance-aware reward shaping.

Abstract

Humanoids deployed in human-centered environments must handle force-interactive tasks, where external contacts introduce unexpected disturbances that disrupt locomotion accuracy and stability. Existing learning-based approaches rely on broad domain randomization, task-specific force objectives, or learning-based force estimators from motion history, each of which compromises accuracy, task transferability, or out-of-distribution (OOD) robustness. We present Analytical Disturbance-Aware Policy Training (ADAPT), a framework that equips humanoid policies with a physically grounded disturbance observer. The core of ADAPT is an analytical whole-body disturbance observer that estimates residual force/torque online with the accessible robot dynamics, without requiring force/torque sensors. Fed directly into the policy, the estimated disturbances give the humanoid an explicit, physics-derived sense of external force/torque that can generalize across diverse unseen scenes. Experiments on a Unitree G1 humanoid show that ADAPT achieves accurate disturbance prediction and stronger robustness than a proprioception-only baseline under torso perturbations, standing pushes, and asymmetric hand payloads, with improved velocity tracking even on OOD disturbances. Moreover, ADAPT enables penalizing inferred disturbances at lower-body joints to encourage lighter locomotion.

Problem

• humanoid locomotion
• external disturbances
• disturbance robustness
• out-of-distribution generalization
• force interaction

Innovation

• disturbance observer
• humanoid locomotion
• sensor-free force estimation
• policy training
• out-of-distribution robustness