Reproduce “Legged Robot State Estimation Using Invariant Neural-Augmented Kalman Filter with a Neural Compensator” 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
Seokju Lee, Hyun-Bin Kim, Kyung-Soo Kim
The authors introduce InNKF, adding a learned correction to a contact-aided invariant EKF for legged-robot state estimation. A temporal convolutional network maps sensor history and the filter estimate to a geometry-constrained base-state correction. In simulated discrete-obstacle terrain, position ATE is 0.1389 m, versus 0.2036 m for an NN-only baseline and 1.0734 m for InEKF; stairs results are not uniformly better.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.