Nonlinear control with linearised models and neural networks

A nonlinear control strategy involving a geometric feedback controller and adaptive approximation of the plant is presented. The plant is approximated by a linearised model and a neural network which approximates the higher order error terms. Online adaptation of the network is performed using steepest descent with a dead zone function. The proposed strategy is applied to two case studies for output tracking of set points. The results show good tracking comparable with utilising the actual model of the plant (usually unknown) and better than that obtained when using the linearised model alone