MaskAdapt: Learning Flexible Motion Adaptation via Mask-Invariant Prior for Physics-Based Characters

TL;DR

MaskAdapt introduces a two-stage residual learning framework for flexible, mask-invariant motion adaptation in physics-based humanoid control, enabling robust and targeted behavior modification.

Contribution

The paper proposes a novel two-stage residual learning approach with a mask-invariant base policy for versatile motion adaptation in humanoid control.

Findings

Robust motion prior stable under missing observations.

Effective multi-part motion composition within a single sequence.

Superior targeted motion adaptation compared to prior methods.

Abstract

We present MaskAdapt, a framework for flexible motion adaptation in physics-based humanoid control. The framework follows a two-stage residual learning paradigm. In the first stage, we train a mask-invariant base policy using stochastic body-part masking and a regularization term that enforces consistent action distributions across masking conditions. This yields a robust motion prior that remains stable under missing observations, anticipating later adaptation in those regions. In the second stage, a residual policy is trained atop the frozen base controller to modify only the targeted body parts while preserving the original behaviors elsewhere. We demonstrate the versatility of this design through two applications: (i) motion composition, where varying masks enable multi-part adaptation within a single sequence, and (ii) text-driven partial goal tracking, where designated body parts follow kinematic targets provided by a pre-trained text-conditioned autoregressive motion generator. Through experiments, MaskAdapt demonstrates strong robustness and adaptability, producing diverse behaviors under masked observations and delivering superior targeted motion adaptation compared to prior work.