Abstract: In the robotic drilling process of large-scale components, workpiece deformation and clamping deviations often lead to discrepancies between actual geometric features and theoretical digital models, making it difficult for hole positioning and normal vector accuracy to meet stringent process requirements. To address this issue, this paper proposes a machining pose error prediction and compensation method based on local feature matching of the workpiece. Within a component matching framework utilizing local node affine transformation, the method integrates point cloud segmentation, node mapping, and iterative optimization strategies to achieve the joint prediction and compensation of target machining position and normal accuracy. Experimental validation was conducted on a cabin structural test specimen. The results demonstrate that during the prediction phase, the proposed method reduces the hole position deviation from 3.99 mm to 0.22 mm and the normal deviation from 0.86° to 0.09°. In actual robotic drilling tests, the positioning and normal accuracy were improved by 86.7% and 87.9%, respectively. This research provides an effective technical approach for the adaptive robotic machining of large-scale components.

Key words: Robotic adaptive machining, Pose error prediction, Large-scale components, Skeletal node registration, Local affine transformation