PMG: Parameterized Motion Generator for Human-Like Locomotion Control
Chenxi Han, Yuheng Min, Zihao Huang, Ao Hong, Hang Liu, Yi Cheng, Houde Liu
AI summary
Problem
Current whole-body reference-guided humanoid controllers struggle to adapt to high-level commands, rely on massive datasets, and face significant sim-to-real transfer hurdles due to brittle dynamics and proprietary commercial black boxes.
Approach
PMG dynamically interpolates a compact library of parameterized motion clips to generate reference trajectories, enforces kinematic consistency through ground-aware optimization, and pairs this with imitation learning and black-box motor calibration for robust real-world deployment.
Key results
- Generates omnidirectional, full-pose human-like motions from under 10 seconds of parameterized data
- Achieves precise tracking of high-dimensional velocity and posture commands via ground-aware optimization
- Eliminates foot slip and base drift through kinematic constraint enforcement
- Successfully transfers policies to the non-commercial ZERITH Z1 prototype with minimal sim-to-real gap
Why it matters
Offers a practical, reproducible pathway for deploying natural and teleoperatable humanoid control on custom hardware without relying on opaque commercial systems.
Abstract
Recent advances in data-driven reinforcement learning and motion tracking have substantially improved humanoid locomotion, yet critical practical challenges remain. In particular, while low-level motion tracking and trajectory- following controllers are mature, whole-body reference–guided methods are difficult to adapt to higher-level command in- terfaces and diverse task contexts: they require large, high- quality datasets, are brittle across speed and pose regimes, and are sensitive to robot-specific calibration. To address these limitations, we propose the Parameterized Motion Generator (PMG), a real-time motion generator grounded in an analysis of human motion structure that synthesizes reference trajectories using only a compact set of parameterized motion data together with high-dimensional control commands. Combined with an imitation-learning pipeline and an optimization-based sim- to-real motor parameter identification module, we validate the complete approach on our humanoid prototype ZERITH Z1 and show that, within a single integrated system, PMG produces natural, human-like locomotion, responds precisely to high-dimensional control inputs—including VR-based tele- operation—and enables efficient, verifiable sim-to-real transfer. Together, these results establish a practical, experimentally validated pathway toward natural and deployable humanoid control.