Abstract
In this article, the optimal linear filtering problem is investigated for networked control systems with random delays and packet dropouts. The filter may receive the current sensor measurement, the one-step delayed sensor measurement or nothing at each time. When nothing is received by the filter, the predictor of the current sensor measurement which may be delayed or lost is used for compensation. For this measurement compensation model, a new augmented system is established. Based on the established augmented dynamic system, an optimal linear filter dependent on probabilities in the linear minimum sense is developed by an innovation analysis approach. Compared with the filters based on the popular zero-input, hold-input, and modified hold-input compensation strategies, the presented filter has better estimation accuracy. The steady-state filter is also investigated. Simulation examples illustrate the efficiency of the developed filter.
Highlights
R ECENTLY, the linear filtering problem for networked control systems (NCSs) has attracted increasing attention owing to their easy implement and wide applications in many fields [1,2,3,4]
Based on such compensation strategy, the optimal linear estimators are proposed for linear systems with packet dropouts (PDs) and finite-step correlated noises [22]; a Gaussian filter [23] and an unscented Kalman filter [24] are designed for nonlinear systems
From the above two examples, we can see that the developed filtering algorithm has better superiority than the filters with ZI, HI and MHI compensators under the case that one packet at most is only used for filtering update
Summary
R ECENTLY, the linear filtering problem for networked control systems (NCSs) has attracted increasing attention owing to their easy implement and wide applications in many fields [1,2,3,4]. Compared with the above strategies, prediction compensation strategy has more superiority since all the measurement data received previously are used for compensation Based on such compensation strategy, the optimal linear estimators are proposed for linear systems with PDs and finite-step correlated noises [22]; a Gaussian filter [23] and an unscented Kalman filter [24] are designed for nonlinear systems. In this paper, based on the prediction compensation strategy, we will design a LF dependent on probabilities for systems with one-step RD and PDs. It should be point out that the same problem has been studied in our previous work [27], their algorithms are essentially different. Main contributions of this paper include: (a) for systems with one-step RD and PDs, an optimal measurement compensation model is established when no packet is received by the filter.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
