Reproduce “Attention-Based Neural-Augmented Kalman Filter for Legged Robot State Estimation” 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, Kyung-Soo Kim
AttenNKF estimates legged-robot state during foot slip by adding a learned, slip-conditioned correction after an InEKF update. It derives a severity cue from foot motion and uses cross-attention to condition correction on filter history; the filter covariance and recursion remain unchanged. On gravel, mean relative position error over 5 m was 0.048 m versus 0.700 m for SR after real fine-tuning.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.