From classical to Networked Control: Retrofitting the Concept of Smith Predictors

Abstract: Filtered Smith predictors are well established for controlling linear plants with constant time delays. Apart from this classical application scenario, they are also employed within networked control loops, where the measurements are sent in separate packets over a transmission channel that is subject to time-varying delays. However, no stability guarantees can be given in this case. The present paper illustrates that the time-varying delays as well as the packetized character of the transmissions have to be taken into account for stability analysis. Hence, three network protocols, which use different packet selection and hold mechanisms, are considered. Criteria for robust stability of the networked feedback loop are given. They are based on the small gain theorem and allow a computationally inexpensive way to check stability for the case with bounded packet delays. Simulation examples provide insight into the presented approach and show why the inclusion of the time-varying packetized character of the network transmissions is vital for stability analysis.