A neural network assisted computed torque method for manipulator tracking control problems

Based on the structure of the computed torque method, this paper develops a neural control method for manipulator trajectory tracking problems. Compared with conventional approaches, the proposed method has the following novel features. First, by using a neural network as the plant model, the proposed method can handle structural as well as unstructural uncertainties. Secondly, with a simple compensation scheme, the proposed approach can effectively avoid tracking performance degradation caused by unmodelled dynamics and/or neural network learning errors. Computer simulation results are given to demonstrate the feasibility of the proposed control method.