Recent advances in text-to-motion generation enable realistic human-like motions directly from natural language. However, translating these motions into physically executable motions for humanoid robots remains challenging due to significant embodiment differences and physical constraints. Existing methods primarily rely on reinforcement learning (RL) without addressing initial kinematic infeasibility. This often leads to unstable robot behaviors. To overcome this limitation, we introduce Contact-aware motion Refinement (CoRe), a fully automated pipeline consisting of human motion generation from text, robot-specific retargeting, optimization-based motion refinement, and a subsequent RL phase enhanced by contact-aware rewards. This integrated approach mitigates common motion artifacts such as foot sliding, unnatural floating, and excessive joint accelerations prior to RL training, thereby improving overall motion stability and physical plausibility. We validate our pipeline across diverse humanoid platforms without task-specific tuning or dynamic-level optimization. Results demonstrate effective sim-to-real transferability in various scenarios, from simple upper-body gestures to complex whole-body locomotion tasks.
(A) Our pipeline begins with text-to-motion generation from natural language, followed by robot-specific retargeting and Contact-aware motion Refinement (CoRe), which includes detecting stable contact segments, optimizing trajectories under contact constraints, adjusting feet orientations, and handling collisions.
(B) The refined motion and extracted contact segments are utilized in physics-based imitation learning, where a reinforcement learning policy is trained with contact-aware rewards. This enables robust sim-to-real deployment, ensuring reliable and safe execution of robot motions corresponding to given text instructions.
1) Contact Segment Detection : Identifying reliable foot-ground contacts (\(\mathcal{C}_f\)) by analyzing toe trajectories.
2) Contact-Constrained Trajectory Optimization : Refining trajectories to eliminate foot sliding and floating, ensuring stable ground interactions and smooth base motion.
3) Feet Orientation Adjustment : Optimizing foot yaw orientation to maintain natural and stable foot positioning during contacts.
4) Collision-handling and Smoothing : Resolving self-collisions through targeted position adjustments and smoothing trajectories to prevent abrupt changes.
Success rate measures stability and motion completeness (≥90% duration without falling).
| Method | Simple Stationary | Walking | Complex (Self-Collision) |
|---|---|---|---|
| Ours | 100% | 73.3% | 66.7% |
| Words into Action | 100% | 23.3% | 6.7% |
| LAGOON | 100% | 26.7% | 20.0% |
Our method notably outperforms baselines, particularly in challenging scenarios involving locomotion and self-collisions.
@inproceedings{jeong2025core,
title = {CoRe: A Hybrid Approach of Contact-aware Optimization and Learning for Humanoid Robot Motions},
author = {Jeong, Taemoon and Chai, Yoonbyung and Choi, Sol and Bak, Jaewan and Kim, Chanwoo and Yoon, Jihwan and Lee, Yisoo and Lee, Kyungjae and Kim, Joohyung and Choi, Sungjoon},
booktitle = {Proceedings of the IEEE-RAS International Conference on Humanoid Robots (Humanoids)},
year = {2025},
note = {Accepted}
}