Kimodo-Policy: Learning Kinematic Action Models for Humanoid Loco-Manipulation
Kimodo-Policy: Learning Kinematic Action Models for Humanoid Loco-Manipulation
This paper presents Kimodo-Policy, an approach that treats whole-body motion as a direct pretraining target using a frozen text-to-motion backbone and lightweight visual conditioning. The method achieves a 72.14% overall success rate on the multi-task HumanoidArena benchmark and demonstrates competitive performance on SIMPLE and six real-world tasks with 60% mean success across eight tasks, utilizing 678M parameters (92M trainable).
The authors introduce Kimodo-Policy, a humanoid control method that uses a frozen text-to-motion backbone with lightweight visual conditioning to achieve 72.14% success on the HumanoidArena benchmark and 60% mean success across eight real-world tasks.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
- paper: not_reported
- code: not_reported
- weights: not_reported
- dataset: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
visual conditioning. A root-body-hand hierarchy coordinates locomotion and manipulation, while motion-history conditioning and repeated action-chunk prediction bridge motion synthesis and closed-loop control. Following large-scale loco-manipulation pretraining, the policy achieves the highest overall success among compared policies on multi-task HumanoidArena (72.14%), competitive performance on SIMPLE, and 60% mean
Open source S4
success across eight real-world tasks, with 678M parameters (92M trainable). Ablations support the contribution of the pretrained motion backbone. Together, these results provide initial evidence for motion generation as a complementary foundation for humanoid action learning.
Open source S5
Source：arXiv Robotics — research abstracts · arxiv.org