Adaptive Admittance Control for Robotic Bolt Grinding

Abstract

To address time-varying environmental stiffness, wheel-surface position drift, and repeated contact search in continuous bolt grinding, an adaptive admittance constant-force control method based on online environmental parameter estimation is proposed. A variable-forgetting-factor recursive least-squares estimator is used to identify the equivalent environmental stiffness and wheel-surface position online. The estimated stiffness is then used to adapt the virtual stiffness of the admittance model, while the damping ratio is regulated according to the force error. In addition, terminal-window position memory is introduced to reuse the estimated wheel-surface position between adjacent bolts. Simulation and robotic grinding experiments show that, compared with the fixed-parameter admittance baseline, the proposed Predefined-Performance Adaptive Perturbation Attenuation (PF-APA) strategy reduces the contact-stage overshoot by 80.51% and shortens the ±2% settling time by 73.03%. In grinding experiments, the proposed method reduces the peak force overshoot from 4.568 N to 0.894 N and shortens the maximum instability duration from 0.006 s to 0.002 s. This study improves force stability and contact efficiency in continuous robotic bolt grinding.