Table of Contents
|2 Uncertain Geometric Information||
|3 Segment-based Representation of Indoor Environments||
|4 Detecting High-Level Features by Multisensor Fusion||
|5 The First-Location Problem||
|6 Simultaneous Localization and Map Building||
|A Transformations and Jacobian Matrices in 2D||
|B Operations with Uncertain Locations||
|C Geometric Relations||
|D Experimental Equipment||
|The present work is a revision of the
doctoral dissertation of J. A. Castellanos presented in the Department
of Computer Science and Systems Engineering of the University of Zaragoza
(Spain) in May 1998.
The work is structured in six main chapters which progressively discuss the contributions of our research introduced in chapter 1. Thus, chapter 2 briefly describes the Symmetries and Perturbations Model (SPmodel), the geometric entities used throughout the work and some results about suboptimal estimation techniques, describing both a recursive formulation and a batch formulation. Chapter 3 deals with the construction of a segment-based representation of the local environment of the mobile robot by using a laser rangefinder sensor. In chapter 4 we semantically upgrade the segment-based representation by using higher level features obtained by multisensor fusion. A monocular vision system provides redundant information about the environment which is combined with the information gathered by the laser rangefinder to increase the robustness and reliability of features from early stages of the processing. Chapter 5 presents the first-location problem as a matching between observed features, both monosensorial features and multisensorial features, and an apriori hand-made model of the navigation area. Finally, chapter 6 describes our contributions to the problem of simultaneous mobile robot localization and map building, by introducing the concept of Symmetries and Perturbations Map (SPmap). Experimental results are described throughout the work to verify the applicability of the theoretical results to the real navigation of a mobile robot in a human-made indoor environment.