Methods to design robust controllers against nonlinear and multiple uncertainties by use of neural networks

In general, both unstructured and structured uncertainties coexist in an uncertain system. Therefore methods in which either type of uncertainties can be treated are not enough, and it is not assured that designed controllers can stabilize the uncertain system. Moreover, methods to design controllers which have robust performance even if the system is mutated are also required. In the paper, methods to design robust controllers for a nonlinear system where both unstructured and structured uncertainties, that is multiple uncertainties, exist. A neural network is effective in designing such a controller, and robustness can be drawn from the training. Effectiveness of proposed methods are shown by some numerical simulations.