Mobile Robot Localization Using Pattern Classification Techniques

This thesis describes a methodology for coarse position estimation of a mobile robot within an indoor setting. The approach is divided into two phases: exploration and navigation. In the exploration phase, the mobile robot is allowed to sense the environment for the purpose of mapping and thus "learning" its unknown surroundings. In the navigation phase, the robot senses the surroundings and compares this information with its learned maps for the purpose of locating itself in the workspace. We solve the task of coarse-level mobile robot localization via pattern classification of grid-based maps of important or interesting workspace regions. Using datasets representing 10 different rooms and doorways, we estimate a 94% recognition rate of the rooms and a 98% recognition rate of the doorways. We conclude that coarse position estimation is possible through classification of grid-based maps in indoor..