<p>Robot-assisted polishing methods have attracted much interest as a replacement for specialized and costly computer numerical control (CNC) polishing tools to achieve high-precision polishing of optical components. This study theoretically and experimentally investigates the surface generation mechanism in the robot-assisted polishing of optical glass. The contact pressure distribution between the polishing head and workpiece is simulated using finite-element software based on the Hertz contact theory and geometric models. A polishing model is proposed based on the Preston theory and experimental results, and the Preston coefficient <i>k</i> of the model is fitted. Subsequently, the local speed of the polishing head in the contact area is analyzed to provide an accurate evaluation of the material removal profile. To realize optimal polishing conditions, polishing experiments with various parameters are conducted in one area without feeding the workpiece. The results of robot-assisted polishing experiments on the entire surface of the workpiece showed mirror surface quality and improved profile accuracy. The obtained results not only clarify the surface generation mechanism in robot-assisted polishing but also highlight the need for further studies of high-precision and low-cost polishing systems of optical components, contributing to the widespread availability of optical devices in industrial applications.</p>