Dottorato Di Ricerca in Ingegneria Informatica Ed Elettronica Industriali - Xv Ciclo Omnidirectional Vision for Mobile Robots

This dissertation describes the work done by the author in the period of his Ph.D. in the field of omnidirectional vision. After a general presentation on omnidirectional vision, the dissertation focuses on omnidirectional vision for mobile robotics. Each chapter of the dissertation covers a different topic in omnidirectional vision for mobile robotics. The three main chapters describe the evolution of an omnidirectional vision system to be mounted on a mobile robot. First, a new simple algorithm used to produce low-cost omnidirectional mirrors with custom profiles is presented. Second, a new method of hierarchical localisation is described. Hierarchical localisation is defined as the determination of the robot’s position with different accuracies depending on the environment structure. Third, the implementation of the Spatial Semantic Hierarchy is described in detail. The Spatial Semantic Hierarchy of B. Kuipers has been previously implemented only on simulated robots or on robots with sonar in very simple environments. The described implementation deals with a real robot, in a real-world environment, equipped with an omnidirectional vision sensor only. The last chapter describes an extension of omnidirectional vision from single robot domain to multi-robot domain. The aim is to build an Omnidirectional Distributed Vision System. In a Distributed Vision System, the different sensors are networked. The interaction and the communication between the sensors enables intelligent behaviours not achievable with a single sensor. This last section should be considered as an overview of the current work of the author, in which preliminary results and preliminary experiments are reported.