On the application of block transmissions for improving control over lossy networks

Abstract

Over the last few years the problem of optimal estimation and control over lossy networks has been studied in detail. However, most of such works assume that in the absence of a new actuation value, the actuator either outputs zero or the previous value. Some other extensions have been proposed, but they present similar limitations. This paper shows that this method is not optimal and presents a new approach, in which the controller sends a number of predictions of future actuation values, which are outputted when the actuator does not receive a new message. Under this new assumption it is shown that the controller/estimator herein presented performs equal to or better than any previously introduced controller, being the equality verified only in the absence of communication errors. A modified Kalman filter as well as a specific optimal controller for this novel actuation model are also presented.