Neural networks for the adaptive control of nonlinear systems

Neural networks are used in an indirect model reference adaptive control technique to identify, then control, a nonlinear system. First, a neural network is used to identify the system. Then, this identifier is used in place of the nonlinear system to adjust a neural network controller. The effect of model mismatch on system convergence and stability is explored. Examples include the well-known inverted pendulum problem. It is shown that for many cases this technique does a fairly good job of controlling the system