Approximate linearization of nonlinear systems: a neural network approach

Recent researches show that neural networks have the ability to approximate a function as well as its derivatives. This result offers a promising opportunity to introduce neural network theory into nonlinear system control. In this paper a novel method of approximate nonlinear system linearization with neural networks is proposed. The network approximator is designed to integrate the involutive equation of a nonlinear system no matter whether the integrability condition is satisfied or not. Simulation results show that this method is feasible.