Full-observability analysis and implementation of the general SLAM model
Ae. Souici, M. Courdesses, Abdelaziz Ouldali, Raja Chatila · International Journal of Systems Science · 2011
Simultaneous localisation and mapping (SLAM) problem is a non-linear system with dynamic state and measurement dimensions. The full observability of such system was ignored, in spite of the fact that system observability is a fundamental aspect in any state estimation problem. In this article, we present a full observability analysis of the general SLAM model. We show that known landmarks (anchor) solution does not guarantee full observability. Furthermore, we prove that to make the general SLAM model fully observable, a combination of known landmarks and invariant metrics are needed. Moreover, we propose a solution to implement a fully observable SLAM model based on mature landmark and virtual observation concepts. Simulations and experimental results are presented demonstrating the validity of the solutions in real world.