Reproduce “TOP: Time Optimization Policy for Stable and Accurate Standing Manipulation with Humanoid Robots” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Zhenghan Chen, Haocheng Xu, Haodong Zhang, Liang Zhang, He Li, Dongqi Wang, Jiyu Yu, Yifei Yang and 2 more
Researchers from Zhejiang University developed the Time Optimization Policy (TOP) framework for humanoid robots, enabling stable and accurate standing manipulation by adaptively adjusting the speed of upper-body motions. The system achieved a 95.30% success rate on unseen motion datasets and demonstrated improved precision and stability in real-world tasks.
Key insights
- Representing upper-body motions using a VAE provides a structured, generalized latent space that enhances coordination awareness for the lower-body controller.
- Adaptively optimizing the time trajectory of upper-body motions is crucial for balancing stability, precision, and time efficiency, mitigating the dynamic burden on the lower body.
- A decoupled training approach, separating the motion prior, lower-body balance policy, and time optimization policy, allows for modular development and improved performance.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.