Abstract:The accuracy of moving horizon estimation (MHE) suffers significantly in the presence of measurement outliers. Existing methods address this issue by treating measurements leading to large MHE cost function values as outliers, which are subsequently discarded. This strategy, achieved through solving combinatorial optimization problems, is confined to linear systems to guarantee computational tractability and stability. Contrasting these heuristic solutions, our work reexamines MHE from a Bayesian perspective, unveils the fundamental issue of its lack of robustness: MHE's sensitivity to outliers results from its reliance on the Kullback-Leibler (KL) divergence, where both outliers and inliers are equally considered. To tackle this problem, we propose a robust Bayesian inference framework for MHE, integrating a robust divergence measure to reduce the impact of outliers. In particular, the proposed approach prioritizes the fitting of uncontaminated data and lowers the weight of contaminated ones, instead of directly discarding all potentially contaminated measurements, which may lead to undesirable removal of uncontaminated data. A tuning parameter is incorporated into the framework to adjust the robustness degree to outliers. Notably, the classical MHE can be interpreted as a special case of the proposed approach as the parameter converges to zero. In addition, our method involves only minor modification to the classical MHE stage cost, thus avoiding the high computational complexity associated with previous outlier-robust methods and inherently suitable for nonlinear systems. Most importantly, our method provides robustness and stability guarantees, which are often missing in other outlier-robust Bayes filters. The effectiveness of the proposed method is demonstrated on simulations subject to outliers following different distributions, as well as on physical experiment data.

Trustworthy Regularized Huber Regression for Outlier Detection

Robust Estimation and Shrinkage in Ultrahigh Dimensional Expectile Regression with Heavy Tails and Variance Heterogeneity

Robust Algorithm for Attack Detection Based on Time‐varying Hidden Markov Model Subject to Outliers

Human-in-the-loop Outlier Detection.

A Robust Learning Algorithm for Regression Models Using Distributionally Robust Optimization under the Wasserstein Metric

Outlier detection using distributionally robust optimization under the Wasserstein metric

A New Outlier Detection Method Considering Outliers As Model Errors

A Statistical Learning Assessment of Huber Regression

A Robust Learning Approach for Regression Models Based on Distributionally Robust Optimization.

Similarity Based Robust Probability Latent Variable Regression Model and Its Kernel Extension for Process Monitoring

Outlier Detection Using Generative Models with Theoretical Performance Guarantees

A Distributionally Robust Optimization Approach for Outlier Detection *

On Regularization Based Twin Support Vector Regression with Huber Loss

Functional Adaptive Huber Linear Regression

Correlation-Based Robust Linear Regression with Iterative Outlier Removal

Provable Self-Representation Based Outlier Detection in a Union of Subspaces

Robust Bayesian Inference for Moving Horizon Estimation

Robust Fused Lasso Penalized Huber Regression with Nonasymptotic Property and Implementation Studies

ERM and RERM are optimal estimators for regression problems when malicious outliers corrupt the labels

DETECTION OF MULTIPLE OUTLIERS BY RANDOM ROBUST TESTING

Robust Multi-Kernel Nearest Neighborhood for Outlier Detection