An Extended Kalman Filter with a Computed Mean Square Error Bound

TL;DR

This paper introduces a bound-based extended Kalman filter (BEKF) that computes a valid mean square error bound during recursive estimation, improving reliability for non-linear systems.

Contribution

The paper presents a novel recursive filter that integrates a computed mean square error bound, extending the conventional EKF to provide guaranteed error bounds.

Findings

Reduces to standard Kalman filter for linear systems

Provides a valid mean square error bound during estimation

Applicable to polynomial and general non-linear systems

Abstract

The paper proposes a new recursive filter for non-linear systems that inherently computes a valid bound on the mean square estimation error. The proposed filter, bound based extended Kalman, (BEKF) is in the form of an extended Kalman filter. The main difference of the proposed filter from the conventional extended Kalman filter is in the use of a computed mean square error bound matrix, to calculate the filter gain, and to serve as bound on the actual mean square error. The paper shows that when the system is linear the proposed filtering algorithm reduces to the conventional Kalman filter. The theory presented in the paper is applicable to a wide class of systems, but if the system is polynomial, then the recently developed theory of positive polynomials considerably simplifies the filter's implementation.