Decentralized adaptive neural network state and output feedback control of a class of interconnected nonlinear discrete-time systems
In this paper, novel decentralized controllers are introduced for a class of nonlinear interconnected discrete-time systems in an affine form with unknown internal subsystem and interconnection dynamics. First under the assumption that the state vector of the local subsystem is only measurable, a single neural network (NN)-based decentralized tracking controller is introduced to overcome the unknown internal dynamics as well as the control gain matrix of each subsystem. The NN weights are tuned online by using a novel update law, and thus, no offline training is employed. By using Lyapunov techniques, all subsystems signals are shown to be uniformly ultimately bounded (UUB). Next, the tracking problem is considered by using output feedback via a nonlinear NN observer. Lyapunov techniques demonstrate that the subsystems states, NN weight estimation errors, and state estimation errors are all UUB. Simulation results are provided on interconnected nonlinear discrete-time systems in affine form and on a power system with excitation control to show the effectiveness of the approach. © 2012 AACC American Automatic Control Council).
Publication Source (Journal or Book title)
Proceedings of the American Control Conference
Mehraeen, S., & Jagannathan, S. (2012). Decentralized adaptive neural network state and output feedback control of a class of interconnected nonlinear discrete-time systems. Proceedings of the American Control Conference, 6406-6411. https://doi.org/10.1109/acc.2012.6315493