Comparison of output feedback controls using ANN for mechanical systems

Two adaptive output feedback control schemes are presented for output tracking of a class of mechanical systems. In the first scheme Gaussian RBF neural networks are used to adaptively compensate for the plant´s nonlinearities. The networks weights are adapted by using a Lyapunov-based design. A parameter projection method and high-gain observer are used in this method to achieve semi-global uniform ultimate boundedness. In the second scheme, the structure of mechanical dynamic equation is considered, and the same Gaussian RBF neural networks are used to approximate the energy function ´K´ and ´U´, in which we need not to consider the system´s size, and it is shown that better tracking results are achieved