Search papers, labs, and topics across Lattice.
This paper introduces Context-Aware Motion Priors (CMP), a framework that tailors motion guidance to specific task contexts without requiring manual skill labels or dataset partitioning. By leveraging high-advantage policy rollouts and a demonstration-based objective, CMP effectively learns context-motion compatibility and reweights reference supervision for training a context-conditioned adapter. The approach significantly enhances task performance and sample efficiency across five humanoid control tasks, demonstrating the importance of context in humanoid policy learning.
Adapting motion priors to specific task contexts can dramatically enhance humanoid control performance and efficiency, outperforming traditional uniform approaches.
Motion priors provide powerful guidance for learning naturalistic humanoid behaviors. However, existing methods typically learn a general, task-agnostic prior from the entire reference dataset and apply it uniformly throughout policy training. As a result, the prior cannot distinguish which reference motions are relevant to the current task context, potentially providing irrelevant or conflicting guidance. We present Context-Aware Motion Priors (CMP), a framework that adapts a general motion prior to the current task context without manual skill labels, dataset partitioning, or a separate skill discovery stage. Specifically, CMP learns context-motion compatibility using high-advantage policy rollouts, while a demonstration-based objective keeps the learned relevance grounded in the reference distribution. The resulting relevance scores reweight reference supervision for training a lightweight context-conditioned adapter. To evaluate the effectiveness and generality of CMP, we instantiate it with both Adversarial Motion Priors and Score-Matching Motion Priors. Across five humanoid control tasks, CMP consistently improves task performance and sample efficiency, learns meaningful context-motion alignment, and remains robust to imbalanced reference distributions. These results show that adapting motion priors to task contexts provides more relevant guidance for humanoid policy learning.
