Abstract: To improve the absolute positioning accuracy of industrial robots in automatic drilling of aircraft parts, an optimized compensation method based on extreme learning machine (ELM) algorithm is proposed to establish the relationship between the nominal position and the actual position of the robot flange center. Firstly, the error variation of the absolute positioning error of the robot in the axial directions of the robot coordinate system is analyzed based on the spatial mesh sampling method, and the feasibility of modeling and compensating the positioning error is verified. Then, the error compensation model based on ELM algorithm is established, and the number of hidden layer neurons is optimized for error model training. Experimental results indicate that the absolute positioning error shows different characteristics in different axial directions. The robot absolute positioning error before compensation is 0.29 mm~0.58 mm, and the average value is 0.41 mm. After compensation, it is reduced to 0.04 mm~0.32 mm, and the average value is 0.18 mm. In addition, the compensation modeling based on ELM algorithm is of high speed and good generalization performance.