Reproduce “OCELOT: Odometry and Contact Estimation for Legged 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
Emre Girgin, Cagri Kilic
OCELOT, a proprioceptive-only odometry method for quadruped robots, integrates learned dynamics-based pseudo-measurements into an Invariant Extended Kalman Filter (IEKF) using calibrated set-coverage statements rather than a Gaussian noise assumption. This method maintains estimator consistency under non-Gaussian errors and achieves competitive accuracy with perception-based systems in challenging real-world scenarios.
Key insights
- Robust contact detection for legged robots requires fusing complementary information, such as physical force and kinematic stationarity, to effectively identify and reject slippage events.
- Dynamically quantifying measurement uncertainty based on a continuous contact quality score allows an ESEKF to adaptively trust or down-weight ZUPT corrections, significantly improving filter consistency and accuracy.
- Adaptive thresholds for force-based contact detection, derived from Gaussian Mixture Models, enable the system to adjust to varying terrain characteristics without manual tuning, enhancing robustness.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.