Outlier-robust zonotope set-membership filter for discrete-time nonlinear system

Transient sensor malfunction, improper replacement measures, and large non-Gaussian noise can cause data outliers in real-world scenarios. In this article, an outlier-robust zonotope set-membership filter is proposed for uncertain nonlinear systems susceptible to outliers. In order to linearize nonlinear systems, a linearization method based on Stirling's interpolation formula is proposed. Then, an outlier detector utilizing the linearization error bounds, the prediction domain, and the output feasible domain is constructed. Based on the linearized system, a tight strip construction method is given to reduce the conservatism of the estimated enclosure. Moreover, a prediction-correction-backward reconstruction filter structure is designed to improve estimation accuracy and reduce conservatism. The proposed structure fuses historical data to attenuate the impact of uncertainty on the system. The robustness, guarantee, and convergence of the proposed filter are investigated. Finally, the effectiveness of the proposed approach is evaluated by two simulation examples.

Automated summary

In this article, an outlier-robust zonotope set-membership filter is proposed for uncertain nonlinear systems susceptible to outliers. A linearization method based on Stirling's interpolation formula is proposed to linearize nonlinear systems, and an outlier detector utilizing the linearization error bounds, the prediction domain, and the output feasible domain is constructed. The effectiveness of the proposed approach is evaluated by two simulation examples.