Legged Robot Estimation - A Decoupled Approach via Moving Horizon Estimation and Invariant Filters

Abstract

In this abstract, we present a modular state estima- tion framework for legged robots that decouples floating-base orientation estimation—handled via an invariant EKF—from the estimation of physical quantities such as contact forces and inertial parameters using a Moving Horizon Estimation (MHE) approach. This design leverages system symmetries and naturally incorporates robot dynamics and deterministic constraints. Build- ing on prior hardware validation, we further integrate recent advances in differential dynamics and numerical optimization to model contact interactions within the estimation process.

Jiarong Kang 康家荣

PhD Student

Jiarong joined the lab in the Fall of 2023 as a PhD student.

Xiaobin Xiong 熊晓滨

Assistant Professor

Prof. Xiong is a full-stack roboticist who develop rigorous theories for practical applications.