Dynamic System State Estimation with a Resilience to Observation Data Anomalies
Keywords:adaptation, abnormal errors, Kalman filter, pseudo-Bayesian estimates
In practical situations, sudden changes in an object's dynamic properties can cause data distortion and random inaccuracies in observation results. These changes often occur due to malfunctions or failures in individual nodes or subsystems. This paper focuses on developing a filter that generates control object state estimates that can withstand fault actions in the measurement subsystem. To achieve this, we modify the observation channel model to account for varying accuracy levels, including abrupt abnormal errors. Our filter synthesis uses Kalman optimal filtering theory methods within the Bayesian framework. The synthesis consists of filtering algorithms that produce a final state vector estimate as a linear combination of model-matched pseudo-Bayesian estimates, weighted by specific coefficients. We provide a rationale for the existence of such estimates and offer an accuracy assessment. We pay particular attention to robust estimators, obtained by simplifying either the structure of the optimal estimator or the weighted coefficient calculation process. To address the a priori uncertainty of the observation channel's anomalous error probabilities, we propose an adaptive estimation algorithm based on observation outcomes. The effectiveness of our synthesized structures is demonstrated through an illustrative example, and we conduct a comparative analysis of their accuracy and associated computational complexity.
How to Cite
Copyright (c) 2024 Taras Maksymyuk, et al.
This work is licensed under a Creative Commons Attribution 4.0 International License.