Driving · 2026-11-10
Kimodo-Policy: Learning Kinematic Action Models for Humanoid Loco-Manipulation
Yunheng Wang, Jiaxi Zhang, Wei Lu, Yuetong Fang, Taowen Wang, Yixiao Feng, Shuning Zhang, Junzhe Xu, Yichi Wang, Zizhao Yuan, Kailun Huang, Lusong Li, Mingqian Zhou, Zecui Zeng, Renjing Xu
AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.
TL;DR
Kimodo-Policy introduces Kinematic Action Models (KAMs), treating whole-body motion as a direct pretraining target using a pretrained motion generator as the action-generation backbone. The policy freezes a Kimodo text-to-motion backbone and adds lightweight ControlNet-based visual conditioning, with a root–body–hand architecture and motion-history conditioning for closed-loop loco-manipulation.
Why it matters
The authors propose motion-generation pretraining as a complementary foundation for humanoid action learning, aiming to transfer temporal and cross-body coordination structure directly into closed-loop robot control. Kimodo-Policy is an initial instantiation that combines a frozen motion backbone with visual adaptation and repeated action-chunk prediction for real-world humanoid loco-manipulation.
Method
- Introduce Kinematic Action Models (KAMs): use a pretrained motion generator as the primary action-generation backbone, with task conditioning, visual observations, and execution feedback to adapt motion priors for robot control.
- Build Kimodo-Policy: freeze Kimodo’s text-to-motion backbone; add lightweight ControlNet-based visual conditioning via trainable root/body branches; use a root–body–hand (LLM2Vec + DINOv3 + HandNet) architecture.
- Use motion-history conditioning and repeated action-chunk prediction to bridge motion synthesis and closed-loop execution with a whole-body tracking controller.
Abstract (from arXiv)
Humanoid loco-manipulation requires continuous coordination of high-dimensional whole-body motion to realize task objectives. While vision-language-action models and world action models transfer semantic and visual-dynamics priors to robot control, these priors do not necessarily capture the structure of robot motion. We introduce Kinematic Action Models (KAMs), a paradigm that treats whole-body motion as a direct pretraining target and uses a pretrained motion generator as the primary action-generation backbone. KAMs ground learned temporal and cross-body coordination priors through task conditioning, visual observations, and execution feedback. We present Kimodo-Policy as an initial instantiation, combining a frozen Kimodo text-to-motion backbone with lightweight ControlNet-based 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 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.
Related papers
- Coordinating Humanoid Loco-Manipulation through Action-Conditioned Latent Foresight
- Being-M0.7: A Latent World-Action Model for Humanoid Robots
- Action-Aligned Latent Planning for Versatile Humanoid Loco-Manipulation
- Language-Forcing: Advancing Humanoid Loco-Manipulation via Semantic Alignment of Whole-Body World Action Models
- Humanoid World Action Model With Joint State--Action Generation
Publication record
Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.
First announced by arXiv: 2026-11-10 · Indexed by Pub-AI: 2026-11-10 UTC
Kimodo-Policy: Learning Kinematic Action Models for Humanoid Loco-Manipulation
First announced by arXiv: 2026-11-10 · Original record ↗
- arxiv
- 2611.11554
Explicit source versions (1)
- v1 · Date not supplied
- Preprint available
Source date: 2026-11-10 · First detected: 2026-11-10 UTC · Historical baseline
Source evidence (1)
- Official evidence ↗ · RSS guid=oai:arXiv.org:2611.11554v1; arxiv:announce_type=new; item pubDate (official metadata)
Abstract · arxiv
Humanoid loco-manipulation requires continuous coordination of high-dimensional whole-body motion to realize task objectives. While vision-language-action models and world action models transfer semantic and visual-dynamics priors to robot control, these priors do not necessarily capture the structure of robot motion. We introduce Kinematic Action Models (KAMs), a paradigm that treats whole-body motion as a direct pretraining target and uses a pretrained motion generator as the primary action-generation backbone. KAMs ground learned temporal and cross-body coordination priors through task conditioning, visual observations, and execution feedback. We present Kimodo-Policy as an initial instantiation, combining a frozen Kimodo text-to-motion backbone with lightweight ControlNet-based 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 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.
