Estimator Stability Analysis In SLAM

Abstract

This work presents an analysis of the state estimation error dynamics for a linear system within the Kalman filter based approach to Simultaneous Localization and Map Building. Our objective is to demonstrate that such dynamics is marginally stable. The paper also presents the necessary modifications required in the observation model, in order to guarantee zero mean stable error dynamics. Simulations for a one-dimensional robot and a planar vehicle are presented.