In order to make up for the deficiency in traditional industrial robots repeat positioning accuracy such as the openloop test, high cost test and complex test process, a measuring method was put forward that laser sensor testing system combined with ant colony optimization neural network algorithm. Through the fast convergence of the ant colony optimization neural network algorithm, it can give a fast and accurate forecast of robot position in repeated positioning test could be given, and positioning accuracy compensation was made for robot by the prediction results. Verified by test, repetitive positioning accuracy prediction results reach the goal of the target error, which can compensate the robot localization accuracy in repeated positioning test.