A Predictor-Based Framework for Delay Compensation in Networked Closed-Loop Systems

One of the main challenges in networked closed-loop systems is that communication delays can significantly deteriorate system performance and even cause instability. A predictor-based framework was proposed previously to improve the networked closed-loop system performance by compensating constant network delays. The key feature of this prediction scheme is that it does not require a dynamic model of the remote subsystem and has only one design parameter to tune for each predictor to mitigate the communication delays. In this paper, the predictor-based framework is developed further. Specifically, previously the predictor was developed and its stability was analyzed for constant delays only. In this paper, the predictor structure is extended to the varying delay case and its stability is established for the first time for varying delays in a sufficient but not very conservative range by leveraging a Markov chain based approach from the literature. To that end, the key assumption of the approach that varying delays can be modeled using a two-time-scale Markov chain is tested using actual network delay data. Next, to test the performance of the framework on a networked closed-loop system, it is applied to a simulated networked motor-shaft-geartrain system with different constant and varying delay models, as well as a real network. Predictors are shown to be effective in significantly reducing the negative effects caused by delays and greatly improve the fidelity of the networked integration. Finally, stability of the closed-loop system with predictor-based framework is studied with constant delays using a “mixed” small gain and passivity theorem from the literature and it is demonstrated that the predictors can even help stabilize a closed-loop linear time-invariant system when the nonpredictor-based networked system is unstable due to constant delays.