Inverse-dynamics adaptive control: a neural network approach

There is a need to develop robust adaptive control algorithms which can function under increased uncertainty. In this situation it is almost mandatory for the controller to have learning and adaptation features. To meet the above stringent design needs, this paper presents a different technique, inverse-dynamics adaptive control (IDAC), using a neural network approach. Simulation results presented illustrate that the learning of the plant dynamics is achieved during the controlling process, that is, learning and control are unified into a single phase: learning-while-functioning. The use of IDAC for control purposes is rather a direct approach in contrast to the conventional adaptive and learning techniques. Furthermore, the IDAC scheme is independent of the type of plant to be controlled, however, in this paper, only linear plants with parameter uncertainties are considered