Robust state estimation using mixed integer programming

Abstract

This letter describes a robust state estimator based on the solution of a mixed integer program. A tolerance range is associated with each measurement and an estimate is chosen to maximize the number of estimated measurements that remain within tolerance (or equivalently minimize the number of measurements out of tolerance). Some small-scale examples are given which suggest that this approach is robust in the presence of gross errors, is not susceptible to leverage points, and can solve some pathological cases that have previously caused problems for robust estimation algorithms.