Abstract: In this paper, the dynamic characteristics of Stanford arm manipulator is described by a discrete time model, and the adaptive control of Stanford arm is studied in this paper. In the adaptive control of industrial manipulator, there always exits the problem that the great change in parameters always leads the global system unstable. To solve this kind of problem, multiple models are used to cover the uncertainties of system parameters, and a multi-model adaptive controller is set up based on these models. At the same time, "localization" method is introduced to reduce computing time, and the accuracy of multi-model adaptive control algorithm is not lost.