A Bayesian Algorithm for Simultaneous Localisation and Map Building

Abstract

This paper describes a full probabilistic solution to the SimultaneousLocalisation and Mapping (SLAM) problem. Previously, the SLAM problemcould only be solved in real time through the use of the Kalman Filter.This generally restricts the application of SLAM methods to domainswith straight-forward (analytic) environment and sensor models. Inthis paper the Sum-of-Gaussian (SOG) method is used to approximatemore general (arbitrary) probability distributions. This representationpermits the generalizations made possible by particle filter or Monte-Carlomethods, while inheriting the real-time computational advantagesof the Kalman filter. The method is demonstrated by its applicationto sub-sea field data consisting of both sonar and visual observationof near-field landmarks.