Extended Mixed Filtering based on Zonotopic and Gaussian Uncertainties for Discrete-Time Nonlinear Systems

In this paper, we propose a mixed filter for discrete- time nonlinear dynamical systems whose uncertainties are composed of both deterministic and stochastic terms. In practice, such mixed composition of uncertainties may appear when assimilating measured data and approximating models. The unknown-but-bounded terms are here represented by zonotopes, which are efficient representations of centrally symmetric convex polytopes. In turn, the stochastic terms are represented by Gaussian random vectors (GRVs) which address confidence regions with high probability. The proposed state estimator is based on linearized models, using the quasi-linear parameter- varying (LPV) approach. The effectiveness of our proposal is illustrated in two case studies.