PartiGen improves humanoid robot motion generation from text commands
PartiGen: Part-Aware Representations for Language-Conditioned Humanoid Motion Generation
Robotics
Summary
Creating realistic robot movements from instructions is tricky because the robot's body parts must move in a coordinated way and at the right time. PartiGen is a new computer method that understands the robot’s body layout and generates movements part by part, using both the text command and previous motions. It also predicts when a movement should stop, making the actions more natural. Tests show that PartiGen produces better and more consistent robot motions than earlier methods and works well on a real robot.
What this means in practice
- •For roboticists: Generate coherent humanoid robot motions directly from natural language commands using anatomical part-aware modeling.
- •For robot motion planners: Use predicted variable motion durations to improve timing accuracy in robot action sequences.
Authors
Chuxuan Huang, Zhanxiang Cao, Tianyue Zhan, Haoyang Li, Yang Zhang, Liyun Yan, Chenlin Ming, Keqi Zhu, Yizhi Chen, Shengcheng Fu, Zihao Zhou, Xinyi Peng, Kaixuan Zhang, Boyang Xing, Jiawei Bao, Yang Yang, Yue Gao
Abstract
Translating language into humanoid motion requires coordinating body parts, determining motion duration, and producing references suitable for robot execution. We introduce PartiGen, a part-aware latent diffusion framework that incorporates anatomical structure into a shared whole-body motion representation. Part-specific attention processes anatomically masked frame features, and its outputs are fused to combine local limb information with whole-body context. Conditioned on text and motion history, the model generates motion primitives autoregressively. Forward-kinematics supervision encourages kinematic consistency, while a learned end-of-sequence predictor selects a frame-level endpoint after generation. On HumanML3D retargeted to the Unitree G1, PartiGen achieves a segment-level FID of 3.165, a 16.8% reduction over TextOp, and improves R@1 by 1.4 percentage points. In the duration study with 16-frame primitives, endpoint prediction reduces mean absolute duration error from 12.471 to 3.080 frames compared with a fixed-length cutoff. We further evaluate robot execution using a pretrained, frozen GRIT tracker: 409 of 500 generated references (81.8%) pass the posture-based stability screen in MuJoCo. Following simulation screening and manual selection, 26 of 27 attempted motions are judged successful on the real G1.