Reproduce “Online Non-linear Centroidal MPC for Humanoid Robots Payload Carrying with Contact-Stable Force Parametrization” 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
Mohamed Elobaid, Giulio Romualdi, Gabriele Nava, Lorenzo Rapetti, Hosameldin Awadalla Omer Mohamed, Daniele Pucci
Researchers at the Italian Institute of Technology developed an online non-linear centroidal Model Predictive Control (MPC) framework that enables humanoid robots to carry payloads while walking. The system demonstrated stable walking with a 2 kg payload on the iCub3 robot, achieving Center of Mass tracking errors between 0.01m and 0.04m and adapting footstep locations.
Key insights
- Proactively integrating knowledge of the payload's effect into the MPC's prediction and cost function enables the controller to anticipate and compensate for persistent disturbances, leading to smoother and more stable motion.
- Using a contact-stable force parametrization can simplify the structure of non-linear optimization problems for contact-rich tasks, potentially improving computational tractability by embedding feasibility into the control parameters.
- Adaptive footstep placement serves as a primary and effective mechanism for humanoid robots to dynamically compensate for persistent external loads like payloads, maintaining overall balance and stability.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.