PDF Paper ArXiv arXiv Youtube Video Code

CooHOI

Learning Cooperative Human-Object Interaction
with Manipulated Object Dynamics

Jiawei Gao*   Ziqin Wang*   Zeqi Xiao   Jingbo Wang   Tai Wang  
Jinkun Cao   Xiaolin Hu   Si Liu   Jifeng Dai   Jiangmiao Pang†  
*Equal Contributions

NeurIPS 2024 Spotlight

arXiv Paper Video Summary Code

Our Approach

Our framework empowers physically simulated characters to execute multi-agent human-object interaction (HOI) tasks with naturalness and precision.

Our framework employs a two-phase learning paradigm. In the first phase, depicted on the left, we train single-agent carrying skills using the AMP framework. In the second phase, we transfer these single-agent skills to a cooperative context. Notably, we use the dynamics of the object as feedback information and implicit communication channel, as illustrated by the bounding box in the figures.

Demos

Single Agent Skills

Two Agents Skills

Multi Agents Skills

BibTeX

@article{gao2024coohoi,
      title={CooHOI: Learning Cooperative Human-Object Interaction with Manipulated Object Dynamics},
      author={Gao, Jiawei and Wang, Ziqin and Xiao, Zeqi and Wang, Jingbo and Wang, Tai and Cao, Jinkun and Hu, Xiaolin and Liu, Si and Dai, Jifeng and Pang, Jiangmiao},
      journal={arXiv preprint arXiv:2406.14558},
      year={2024}
}