On integration of event-based estimation and robust MPC in a feedback loop

The main purpose of event-based control, if compared to periodic control, is to minimize data transfer or processing power in networked control systems. Current methods have an (implicit) dependency between triggering the events and the control algorithm. To decouple these two, we introduce an event-based state estimator in between the sensor and the controller. The event-based estimator is used to obtain a state estimate with a bounded covariance matrix in the estimation error at every synchronous time instant, under the assumption that the set in the measurement-space that is used for event generation is bounded. The estimation error is then translated into explicit polytopic bounds that are fed into a robust MPC algorithm. We prove that the resulting MPC closed-loop system is input-to-state stable (ISS) to the estimation error. Moreover, whenever the network requirements are satisfied, the controller could explicitly request for an additional measurement in case there is a desire for a better disturbance rejection.