Robust H∞ Filtering for Uncertain Nonlinear Systems using Neural Networks

A full-order robust H
∞
filtering design for a class of uncertain nonlinear systems was investigated. The nonlinearities are modeled by neural-networks and then represented by linear difference inclusions. The uncertainties are described by polytope type. The presented filter is linear time-invariant, which not only guarantees the robust stability of error system but also satisfies a prescribed H
∞
attenuation level for all admissible uncertainties. The sufficient condition for the existence of such robust H
∞
filter is provided in terms of linear matrix inequality. A simulation example is given to illustrate the design procedures.