Reproduce “A Unified MPC Framework for Whole-Body Dynamic Locomotion and Manipulation” 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

Jean-Pierre Sleiman, Farbod Farshidian, Maria Vittoria Minniti, Marco Hutter

Researchers at ETH Zurich developed a real-time Model Predictive Control (MPC) framework that unifies dynamic locomotion and manipulation tasks for mobile manipulators into a single optimal control problem. This framework exploits the robot's whole-body dynamics, including manipulated object dynamics, for agile and coordinated behaviors on hardware.

Key insights

  • Unifying locomotion and manipulation within a single MPC framework enables mobile manipulators to leverage their full kinematic redundancy and whole-body dynamics for more agile and coordinated movements.
  • The robot's arm can function as a 'dynamic tail' to assist balancing during fast locomotion, demonstrating substantial reductions (e.g., 66.8% in base roll) by dynamically shifting the whole-body Center of Mass.
  • Explicitly incorporating manipulated object dynamics into the planner allows the generation of physically consistent motion and force plans for complex interactive tasks like pushing or pulling resistive objects.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.