Outlier-Robust State Estimator :JCMSI 2021

This toolkit addresses the critical challenge of state estimation when observation signals contain intermittent outliers.

MATLAB implementation of the robust state estimation method published in SICE Journal of Control, Measurement, and System Integration (2021). This toolkit addresses the critical challenge of state estimation when observation signals contain intermittent outliers caused by sensor failures, communication errors, or cyber attacks.

The algorithm creates multiple candidate state estimates using measurements from different time steps, then selects the most reliable estimate through median or weighted median operations. This approach automatically rejects outlier-contaminated estimates while maintaining high accuracy during normal operation. The method includes systematic observer gain design based on reachable set analysis and Lyapunov inequalities.

The tool visualizes true versus estimated states, identifies which observer is selected at each time step, tracks outlier occurrences, and analyzes selection frequency statistics. This makes it invaluable for understanding how the median-based selection mechanism provides robustness against sparse outliers without requiring explicit outlier detection algorithms.

Ideal for networked control systems, sensor networks, industrial IoT applications, and any scenario where measurement reliability cannot be guaranteed. The approach is particularly effective when outliers occur infrequently and intermittently, making it suitable for real-world cyber-physical systems.

Reference: H. Okajima, Y. Kaneda, N. Matsunaga, "State estimation method using median of multiple candidates for observation signals including outliers," SICE JCMSI, vol. 14, no. 6, pp. 257-267, 2021. DOI: 10.1080/18824889.2021.1985702

Compatible with MATLAB R2016b and later. No additional toolboxes required.

Basic idea is in Trans. on SICE (2019)

メディアンを用いた外れ値やデータ欠落にロバストな状態推定オブザーバ

Poster Presentation:

https://youtu.be/BxoipHh5l6A?si=tlE_zFImWP9Xbh_k

説明動画

https://youtu.be/G4TUjYiFGfA?si=wovNktyOyf1lH9rp

Blog Articles: State Observer and State Estimation: A Comprehensive Guide (covers Kalman filter, H-infinity filter, multi-rate observer, and MCV observer) | System Identification: From Data to Dynamical Models (multirate system identification for model acquisition) | Model Error Compensator (MEC): Enhance the Robustness (combining identified models with MEC for robust control) | MATLAB File Exchange (Author Profile) | Control Engineering Video Portal (500+ videos)