It is necessary to recalibrate the tool coordinates when replacing tools in the application of industrial robots. Generally, constrained by a fixed point, the tool coordinates are calibrated by the threepoint method. This method will get a bad result when the home position of the robot is not accurate. Based on the principle of fixedpoint constraint in the threepoint method, an improved home position selftuning method for industrial robots was proposed. The calibrated 6-DOF serial robot was used as the experimental object, and the given error was generated by modifying the standard value of its home position and tool coordinate. Based on the constraint relationship of a fixed point in the space, the coordinate transformation model of the robot was established under the condition that the parameters of the robot link were known, and 21 groups of robot joint angles and Cartesian coordinate data were taken as the identification condition parameters.The deviation between the tool coordinate and the home position was regarded as unknown data, and the deviation between the tool coordinate and the home position was calculated by leastsquare iteration algorithm. Compared the experimental result with the given error parameter, the result is basically consistent. The feasibility of the calibration algorithm is proved by comparing the error of the position display value on the teaching device before and after selftuning state.