DoS Attacks-Resistant Fusion Filtering for Nonlinear Rectangular Descriptor Systems with Uncertain Attack Probability

This paper investigates the design problem of a fusion filtering algorithm for nonlinear networked rectangular descriptor systems with multi-source information. A set of independent Bernoulli random variables is introduced to characterize the phenomenon of denial of service (DoS) attacks, where the attack probability consists of a nominal probability and an uncertain term with a known upper bound (UB). Firstly, a full-order transformation method is introduced to transform the rectangular descriptor system into a non-descriptor system, where the local filters are constructed. Then, by employing the elementary inequality, the upper bounds on both the prediction error covariance and the filtering error covariance are derived and characterized by matrix difference equations. The local filter gain is designed by minimizing the obtained UB of the filtering error covariance. Subsequently, based on the covariance intersection fusion rule, a new fusion filtering algorithm is presented for addressing the descriptor system. Lastly, the effectiveness of the proposed algorithm is verified through a simulation experiment.