In order to improve the robustness of parameters identification in the industrial robot calibration, a method for calibration of kinematic parameters is introduced based on the robust estimation. First, the end pose error model is established based on the D-H model, as well as the equivalent weight functions are analyzed. Then the robot base coordinate system and end pose are measured by the laser tracker. Last, the kinematic parameters are identified by the robust least squares method combined with IGG3 weight factor function. The result shows that the absolute positioning RMS error of the robot is decreased from 0.87 mm to 0.21 mm after compensation. The calibration method has higher fitting accuracy and robustness than those of the traditional least square compensation method.