Hole position correction method for robotic drilling based on single reference hole and local surface features

In the field of large thin-walled workpiece manufacturing, hole position correction is critical for enhancing the accuracy of robotic drilling. Currently, adding multiple reference holes is the primary approach to improve the accuracy of hole position correction, which may be impractical or inefficient in engineering applications. High-precision and high-efficiency hole position correction technology remains challenging, especially for workpieces with multi-stage bending and torsional deformations. Therefore, we propose a hole position correction method for robotic drilling based on the single reference hole and local surface features. A new feature vector combining the reference hole, surface normal vector, and height deviation is constructed, which is sensitive to the bending and torsional deformations of the workpiece. Using the local point cloud between the single reference hole and the target hole, a hole position offset prediction network is designed to provide correction values for the target holes. Multiple simulations and experiments are conducted to verify the proposed method. The results demonstrate that the proposed method can use only one reference hole to accurately correct the target hole within a 240 mm x 960 mm area, and the average correction error is approximately 0.10 mm.