Data-Driven Robust Non-Fragile Filtering for Cyber-Physical Systems

Filtering or state estimation plays an important role in the cyber-physical systems (CPSs). This paper aims to solve the data-driven non-fragile filtering problem for the cyber-physical system. Randomly occurring gain variations are considered so as to account for the parameter fluctuations occurring during the filter implementation. The data-driven communication mechanism is utilized to reduce the measurement transmission frequency and save energy for the CPSs. Therefore, a unified H
_∞
filtering framework that combines the data-driven communication mechanism and the non-fragility of filters is constructed. Based on this unified framework, the influence of the simultaneous presence of networked-induced packet dropouts, quantization, randomly occurring nonlinearities and randomly occurring parameter uncertainties in the CPS is investigated. A modified dropouts model is proposed under the data-driven communication mechanism. By utilizing stochastic analysis and Lyapunov functional theory, sufficient conditions guaranteeing the filtering performance are derived. The H
_∞
filter is obtained through the proposed algorithm. Last, a simulation is given to demonstrate the filtering method for CPS in this paper.