Abstract: In the process of automatic assembly of shaft holes, inaccurate positioning of industrial robots will lead to excessive contact force between shaft holes, jamming and low assembly accuracy, which seriously affects the assembly effect. A shaft hole assembly method based on YOLOv11-FoundationPose positioning and force feedback control is proposed. The YOLOv11 algorithm is used to segment the hole part mask in real time, and then the FoundationPose algorithm is used to estimate the position of the hole part in the mask area of the hole part. Then, based on Canny edge detection and Hough circle transform, the geometric center of the hole on the hole part is positioned to provide accurate target hole position for the shaft hole assembly robot. The spiral hole search strategy based on force feedback control further reduces the residual error of hole positioning. At the same time, the admittance control is introduced in the assembly process to adjust the pose of the shaft parts in real time, so as to realize the flexible insertion. The experimental results of shaft hole assembly of three kinds of inclined hole parts of 0°, 15° and 30° show that the success rate of shaft hole assembly of three kinds of inclined hole parts reaches 100%, 98% and 92% respectively under the condition of 0.2 mm shaft hole clearance, which verifies the high robustness and reliability of this method in the unstructured environment with uncertain pose.