Adaptive Hybrid Force/Position Admittance Control With a Velocity-Level Correction for Robotic Grinding Contact Tasks

Abstract Robotic grinding and related surface-contact tasks require smooth contact establishment and stable force regulation during sustained interaction. Fixed-parameter admittance control often cannot meet both requirements, because damping that suppresses initial impact may compromise compliant behavior after contact. This paper presents an applied adaptive hybrid force/position admittance framework for robotic grinding. It integrates velocity-level admittance correction with a projection-constrained adaptive update law and a σ-modification term. The velocity-level correction maps normal force error directly to corrective velocity commands, suiting industrial robots with Cartesian or joint velocity interfaces. The adaptive law regulates virtual inertia and damping online while keeping the estimates inside a physically meaningful feasible set under weak excitation. Lyapunov-based analysis establishes uniform ultimate boundedness of the high-level normal-contact channel under stated local-contact and bounded velocity-servo assumptions. UR5-based grinding experiments under 23 N and 3 N target forces show that the proposed controller reduces force-tracking RMSE and contact-force fluctuation relative to the tested baselines, while also reducing the mean maximum impact force in the evaluated contact-transition stage. The results also show a normal-motion trade-off, since the heuristic adaptive controller produced the smallest mean Z-axis position standard deviation.