Abstract: When milling on the bone curved surface, a robot needs to adjust the angle between the tool and the bone surface while controlling the milling depth to ensure the quality of the operation. For this problem, the vibration model of milling cutter is established firstly, and the influence of the angle between the cutter and the bone surface on milling vibration is discussed. Secondly, the three-axis acceleration sensor fixed on the cutter is used to collect the cutter vibration signals during the milling process, and the fast Fourier transform is used to extract the harmonic amplitude of the cutter rotation frequency. Thirdly, the linear relationship between the amplitude of the first harmonic in the tangential acceleration signal and the depth is fitted through a priori experiment, and the relationship between the amplitude of the first harmonic in the axial acceleration signal and the angle is obtained. It is proved that these two harmonics can be used as the feedback for controlling the milling depth and its angle. Finally, the bone curved surface milling method based on vibration tactile sensation is proposed based on PID (proportional-integral-differential) controller and robot inverse kinematics. Experimental results show that when the milling depth is set to 0.5 mm, the average and standard deviation of the milling depth on the bone curved surface before and after the angle control are 0.455 mm ±0.046 mm and 0.499 mm ±0.028 mm. The proposed method can improve the robot-assisted milling accuracy on the bone curved surface.