Abstract
This paper presents an adaptive hybrid force-position control method for wheel-spoke grinding robots, addressing two critical industry challenges: (1) path-surface mismatch causing localized over/under-grinding, and (2) unstable contact pressure leading to poor surface finish. The proposed method integrates a disturbance observer (DOB) with nonsingular fast terminal sliding mode control (NFTSMC), featuring dual-loop innovation: In the force-control loop, a recurrent neural network (RNN) dynamically adjusts impedance parameters to maintain constant grinding force; in the position-control loop, the DOB-enhanced NFTSMC achieves precise trajectory tracking while rejecting disturbances. Experimental validation on automotive wheel spokes demonstrates superior performance: position tracking error reduced to [Formula: see text] (42% improvement vs. PID control), steady-state force error [Formula: see text]N, and surface roughness Ra[Formula: see text] m (meeting ISO 1302 grinding standards). The control system shows 40% faster convergence than conventional sliding mode methods without singularity issues. Experimental results demonstrate that the proposed adaptive variable impedance hybrid control achieves superior stability and surface quality in robotic spoke grinding tasks.