Robust adaptive nonlinear control using single hidden layer neural networks

We develop an adaptive dynamic inverting controller with guaranteed closed loop stability for partially or completely unknown nonlinear non-affine dynamic systems. We assume full state feedback and no zero dynamics. A single hidden layer neural network is used to approximate the inverse map, and a stable adaptive scheme is used to update online the neural network weights. Stability is guaranteed by introducing a robust adaptive bound. The performance of the adaptive scheme is demonstrated in a tracking task controller design and simulation for the nonlinear Van der Pol oscillator