Nonlinear Network-Induced Time Delay Systems With Stochastic Learning

This paper presents a new control approach for nonlinear network-induced time delay systems by combining online reset control, neural networks, and dynamic Bayesian networks. We use feedback linearization to construct a nominal control for the system then use reset control and a neural network to compensate for errors due to the time delay. Finally, we obtain a stochastic model of the Networked Control System (NCS) using a Dynamic Bayesian Network (DBN) and use it to design a predictive control. We apply our control methodology to a nonlinear inverted pendulum and evaluate its performance through numerical simulations. We also test our approach with real-time experiments on a dc motor-load NCS with wireless communication implemented using a Ubiquitous Sensor Network (USN). Both the simulation and experimental results demonstrate the efficacy of our control methodology.