Sampled-Data Admittance-Based Control for Physical Human–Robot Interaction With Data-Driven Moving Horizon Velocity Estimation

Abstract

This article presents a data-driven scheme that integrates moving horizon estimation (MHE) with sampled-data admittance-based control to stabilize physical human–robot interaction (pHRI). The proposed MHE employs data-driven parameterizations based on a single historical trajectory to reconstruct the interaction dynamics response, as ensured by the extended Willems’ fundamental lemma. To mitigate instability in pHRI systems that may arise from noise-contaminated velocity measurements, we implement an online velocity update mechanism grounded in optimal estimation. The sampled-data approach establishes an appropriate sampling interval, facilitating collaboration between the locally linearized pHRI dynamics and MHE for the generation of data-driven velocity. To validate the effectiveness of the proposed method, we performed numerical simulations and experiments using a three-degrees of freedom (DoF) Phantom Omni haptic manipulator, which demonstrated superior transient and steady-state tracking performance.