Reproduce “Proprioceptive Invariant State Estimation for Humanoid Robots on Non-Inertial Ground” 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

Falak Mandali, Zijian He, Yan Gu

Purdue University researchers developed an Invariant Extended Kalman Filter (InEKF) for humanoid robots, enabling accurate real-time state estimation relative to a non-inertial ground frame using only onboard proprioceptive sensors. This method eliminates the need for external ground motion measurements, achieving faster convergence and higher accuracy for relative base position and velocity compared to prior approaches.

Key insights

  • Accurate inference of dynamic ground motion and relative robot state estimation is achievable using only proprioceptive sensors, by leveraging the kinematic constraints of the stance foot.
  • Observability analysis reveals that relative base velocity is consistently observable, while relative base position observability depends on the angular motion complexity of the stance-foot frame.
  • Full state observability, including base orientation, can be practically recovered by augmenting the robot's base sensing with multiple non-collinear IMUs.

Summary by alphaXiv.

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