Adaptive stabilization of Model-Based Networked Control Systems

In this paper Model Based Networked Control Systems (MB-NCS) are considered and on-line identification of system parameters in state space representation is used to upgrade the model and the controller of the system. The updated model is used to control the real system when feedback information is unavailable. The Extended Kalman Filter (EKF) is analyzed in the context of parameter identification and implemented in the MB-NCS framework. Emphasis is placed on global asymptotic estimators for the case when sensors provide noiseless measurements of the state of a linear system; it can be shown that the identification of parameters in this case is a linear problem, in contrast to the nonlinear combined state- parameter estimation problem. We propose new estimation models that offer better convergence properties than the EKF in this case. This estimation strategy is also applied to the MB- NCS framework resulting in a better usage of the network by allowing longer intervals without need for a measurement update.