Humanoid robots keep gestures and posture looking natural while moving
Beyond Kinematic Tracking: Preserving Expressive Characteristics in Humanoid Whole-Body Control
Robotics
Summary
Humanoid robots often struggle to keep the natural feel of their movements when following commands, focusing too much on exact joint positions rather than the overall motion style. The authors show that just copying positions isn’t enough to keep the expressiveness of actions like how much energy or body lean is involved. They created a system that watches which parts of the body are active and adjusts movements to keep these expressive traits intact. Their approach works better than previous methods in simulations and on a real robot, making robot motions look more natural and expressive.
What this means in practice
- •For robotics engineers: Improve humanoid robot controllers to retain natural gestures and posture during complex motions.
- •For robotics hardware developers: Design improved whole-body control algorithms to enhance motion quality in physical humanoid robots using existing control frameworks.
Authors
Lingfan Bao, Hanshang Zhu, Yan Pan, Jincheng Wang, Jingyang Yu, Tianhu Peng, Yiyang Jia, Dimitrios Kanoulas, Chengxu Zhou
Abstract
Whole-body control plays a crucial role in enabling humanoid robots to execute expressive motions. Faithful execution requires preserving both the intended action and its expressive characteristics. Existing whole-body trackers typically balance joint and keypoint tracking with physical stability across diverse motions, but these objectives do not necessarily prioritize the expressive characteristics of each action. Our analysis shows that lower average positional tracking error does not guarantee better preservation of dynamic and postural motion characteristics. To address this gap, we propose a residual learning framework with compensation objectives targeting amplitude, energy, body expansion, body lean, and pelvis height. To concentrate compensation while limiting interference with the underlying tracking behavior, motion-aware gating allocates correction capacity according to body-part activity, preserving larger correction ranges in more active parts and restricting them elsewhere. For each frozen tracker, we train motion-specific residual policies and distill them into a unified policy shared across motion categories. Simulation and physical experiments on the Unitree G1 across eight motion categories demonstrate superior preservation of dynamic and postural motion characteristics compared with the corresponding baselines.