Finite-time filtering for non-linear stochastic systems with partially known transition jump rates

This study is concerned with the problem of robust finite-time filtering for a class of non-linear Markov jump systems (MJSs) with partially known information on the transition jump rates. The non-linearities in the system are parameterised by multilayer neural networks. Our attention is focused on the design of a mode-dependent full-order H _?? filter to ensure the finite-time boundedness of the filtering error system and a prescribed H _?? attenuation level for all admissible uncertainties and approximation errors of the networks. Sufficient conditions of filtering design are developed in terms of solvability of a set of linear matrix inequalities. A tunnel diode circuit is used to show the effectiveness and potentials of the proposed techniques.