Abstract: A neural controller with a feedforward multilayered network controller is proposed. It is able to learn the inverse of the plant so that it can directly provide the appropriate inputs to the plant according to the feedback input/output data and the desired outputs. Furthrmore, on the basis of an introduced forward model of the Plant with a neural network, the weights of the neural controller can be on-line by backpropagation errors between the actual and desired plant outputs, which maks the neural controller Perform adaptive control.