Search papers, labs, and topics across Lattice.
This study empirically evaluates various design choices in humanoid motion tracking policies using the YAHMP framework, which allows for controlled and reproducible experiments. By systematically varying factors such as motion-command representation and training approaches, the researchers identify which choices significantly impact tracking performance versus those that primarily affect training complexity or actuation effort. The findings reveal that certain modeling decisions lead to improved tracking capabilities and successful zero-shot deployment on the Unitree G1, showcasing effective whole-body motion tracking and balance under perturbations.
Identifying critical design choices in humanoid motion tracking could revolutionize how we approach policy development for robotic motion imitation.
Humanoid general motion tracking requires policies that can follow diverse whole-body references while maintaining balance. Building such policies involves many practical design choices, and their individual effects are often hard to assess. We address this issue with an empirical study of common modeling and training factors used in recent humanoid motion-imitation pipelines. To make the study controlled and reproducible, we developed YAHMP, an open-source modular framework for training, evaluating, and deploying whole-body motion tracking policies on the Unitree G1. Within YAHMP, we define a nominal configuration and compare variants that differ in motion-command representation, observation history, action representation, actuation profile, hand-force randomization during training, and training approach. We evaluate the resulting policies on a test set of retargeted human motions and compare the nominal policy with TWIST2 as an external baseline trained on the same motion set. The results distinguish choices with clear tracking effects from choices that mainly change actuation effort, training complexity, or physical interaction capability. Finally, we deploy YAHMP policies zero-shot on the real Unitree G1, demonstrating diverse whole-body motion tracking, balance under external perturbations, and forceful interaction.
