Abstract: In order to improve the plane positioning accuracy of selective compliance assembly robot arm (SCARA) robot, the method combined compensation strategy based on grid model and the minimum error approximation principle was proposed. Firstly, a simplified model for SCARA robot was established and grid model principle was summarized. Secondly, an initial grid model was constructed by analyzing the location relationship between the first-time reached actual point and the desired point. The starting point of the variable parameter grid model in the next step was constructed by using the minimum distance error approximation method. Finally, the convergence and update direction were determined by the location of the desired point in grid model. The results show that the positioning compensation strategy of visual guidance makes up the phenomenon that the plane positioning accuracy is not high because of the inaccuracy of the model. The accuracy of the positioning compensation strategy is much better than 1mm~3mm of the spatial interpolation compensation method. The proposed method has brief parameter regulation, clearly mobile times and strong industrial application.