A stable neural network-based adaptive control scheme for a class of nonlinear plants

A stable adaptive neural network control scheme, based on that suggested in Boskovic (1997), is designed for a class of second-order unknown nonlinear plants. It is shown that a two-layer neural network, built using a rational activation function, results in a convenient parametrization which enables the generation of stable adaptive laws. The corresponding well-tuned neural network controller results in the stable overall system as demonstrated using a suitably chosen Lyapunov function. It is also shown that this controller can be efficiently combined with an indirect VS controller within the framework of stable multiple model-based adaptive control to achieve the control objective and assure favorable performance of the overall closed-loop system.