While many mobile robots come equipped with ultrasonic range sensing (sonar), accurate map building and position estimation using sonar has been elusive because of the difficulty in interpreting sonar data correctly. This book presents evidence that sonar can in fact fulfil the perception role for the provision of long-term autonomous navigation in a broad class of man-made environments. A new approach to mobile robot navigation is presented that unifies the problems of localization, obstacle detection and map building in a combined multi-target tracking framework. The primary tools of this approach are the Kalman filter and a physically based sonar sensing model. Experimental results with real sonar data demonstrate model-based localization using an a-priori hand-measured map and sub-centimetre accuracy map building for an uncluttered office scene. This book shoud be of particular interest to researchers on mobile robotics, especially potential users of sonar. The approach has greater significance, however, as the issues involved - the choice of representation, the problem of data association and the pursuit of long-tem autonomy - are central to many outstanding problems in robotics and artificial intelligence.