The invention provides a model uncertainty mechanical arm motion control method based on a multi-layer neural network. Firstly, controller design is performed on mechanical arm system model uncertainty according to a nominal model; a mechanical arm system dynamic model with uncertainty is established; a mechanical arm system nominal model is established with an uncertain item caused by external interference factors being considered; a controller is designed for the model uncertain item according to the normal model; the multi-layer neural network is adopted to perform self-adaptive approach onthe model uncertain item; and a mechanical arm system controller is designed based on the multi-layer neural network. According to the model uncertainty mechanical arm motion control method based onthe multi-layer neural network, a good robustness effect on structural uncertainty such as parameters and non-structural uncertainty such as external interference is achieved simultaneously, and it can be guaranteed that the track of the tail end of a mechanical arm and the angles of all joints are well tracked.