Reproduce “Hierarchical Reduced-Order Model Predictive Control for Robust Locomotion on 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

Adrian B. Ghansah, Sergio A. Esteban, Aaron D. Ames

Researchers at Caltech developed a hierarchical Model Predictive Control framework that enables robust, real-time humanoid locomotion by combining a high-level nonlinear MPC for adaptive gait planning with a mid-level linear MPC utilizing a novel Decomposed Single Rigid Body model. The system demonstrated a 36% increase in push recovery success through adaptive step periods and enhanced yaw disturbance rejection via integrated upper-body dynamics, validated on a Unitree G1 robot across diverse terrains.

Key insights

  • A hierarchical architecture effectively balances computationally intensive high-level planning with fast, real-time mid-level execution, crucial for robust humanoid locomotion.
  • Simultaneously optimizing step periods with step lengths provides a powerful mechanism for disturbance rejection, significantly increasing robot stability.
  • Integrating simplified upper-body dynamics into a linear reduced-order model (DSRB) offers an efficient way to enhance stability against yaw disturbances without excessive computational burden.

Summary by alphaXiv.

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