Neual state space model based approximation pole assignment control for a class of unknown nonlinear systems

In this paper, an extended linearized neural state space (ELNSS) model is proposed and used to design an approximate pole assignment control strategy for a class of nonlinear systems. At first, the applicability of the ELNSS model to approximate affine nonlinear systems is studied, where the extended Kalman filter (EKF) algorithm is employed to train the weights of the ELNSS model. It has been shown that such a training algorithm can guarantee the convergence of the network weights. Using the trained weights in the ELNSS model, the design of an approximate pole assignment controller is performed using a state feedback framework. The convergence of the approximate pole assignment adaptive control algorithm is also analyzed.