Reproduce “Mobile Pedipulation for Object Sliding via Hierarchical Control on a Wheeled Bipedal Robot” 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
Yue Qin, Yulun Zhuang, Zelin Shen, Yanran Ding
Researchers at the University of Michigan developed a hierarchical control framework for wheeled bipedal robots to perform dynamic object sliding, termed mobile pedipulation, by integrating a new reduced-order model with NMPC and WBC. The framework enabled a Tron1 robot to slide a 1 kg object laterally from under a desk and scoot a 4 kg object 0.228 meters, while simulations showed the potential to scoot objects up to 23 kg.
Key insights
- Explicitly capturing the hip roll degree of freedom and supporting various wheel-environment contact modes in a reduced-order model is crucial for enabling versatile and dynamic pedipulation behaviors in wheeled bipedal robots.
- Integrating object dynamics directly into the Nonlinear Model Predictive Control (TRBO-NMPC) improves object state tracking accuracy and enhances robustness to uncertainties, such as object mass mismatch, compared to approaches that exclude object dynamics.
- Wheeled bipedal robots can leverage their inherent wheel-environment contact stability for effective non-prehensile object sliding, offering a solution for manipulating objects difficult to grasp or located in confined spaces.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.