FADA: Few-Shot Domain Adaptation via Dynamics Alignment for Humanoid Control

2026-06-26

AI Summary

This work addresses the challenge of dynamics mismatch in humanoid robots caused by variations in terrain, payload, or actuators. The authors propose Planner-IDM, a three-stage framework that first trains an expert policy using privileged information and then distills it into a deployable student policy via DAgger. During deployment, only approximately two minutes of action–state data from the target domain are required to fine-tune an inverse dynamics model (IDM), while the planner remains frozen, enabling highly efficient few-shot domain adaptation. Notably, this approach requires neither optimal demonstrations nor reward signals, substantially reducing adaptation costs. It outperforms context-based adaptation and end-to-end baselines across diverse dynamics-shift scenarios and successfully executes a variety of high-precision whole-body tasks on a real humanoid robot.

Abstract

High-precision humanoid control is limited by target-domain dynamics mismatch, where the same control objective can induce different realized motions under changes in terrain, payload, or actuator response. Existing methods either pursue zero-shot transfer through domain randomization or in-context adaptation without target-domain specialization, or require heavy adaptation pipelines that leverage target-domain data, such as model calibration, residual learning, or policy retraining. In this paper, we present FADA (Few-Shot Domain Adaptation via Dynamics Alignment), a three-stage Planner-Inverse Dynamics Model (Planner-IDM) framework for few-shot adaptation in humanoid control. FADA first trains an oracle policy with privileged information and then distills the oracle behavior into a deployable Planner-IDM student through DAgger. At deployment, FADA freezes the planner and finetunes only the IDM using approximately 2 minutes of target-domain rollouts with standard supervised learning. Rather than requiring optimal demonstrations or rewards, FADA uses the paired actions and observations that are observed during these rollouts as supervision, aligning the IDM's action generation with target-domain dynamics. Experiments show that FADA outperforms both in-context and end-to-end adaptation baselines, improving task performance under dynamics shifts and enabling real humanoid robots to execute diverse high-precision whole-body tasks. Implementation details and qualitative hardware rollout videos are available at https://lecar-lab.github.io/FADA-humanoid/.

Problem

• humanoid control
• domain adaptation
• dynamics mismatch
• few-shot learning

Innovation

• Few-Shot Domain Adaptation
• Dynamics Alignment
• Inverse Dynamics Model
• Humanoid Control
• Behavior Distillation