Evaluation

Contents

In this Chapter, I present an evaluation of my approach to robot mapping. Evaluation is performed in experiments using simulated and real-world sensor data. In a first study, I present results from examining self-localization performance. Self-localization performance is essentially governed by the performance of correspondence determination, thus, providing good insight in results achieved as regards one of my central objectives of developing a robust approach to correspondence determination on the basis of shape analysis. Experiments cover the standard task of relating observations to the robot’s internal map. In addition, one setup employs an external, coarse map, to which the observations of the robot need to be related. This setting is included to evaluate achievements as regards my research goal to advance robot navigation towards utilizing externally supplied maps.

In a second study, mapping performance is investigated. This analyzes the robustness to localization in partially unknown environments and the reliable interplay of all functional components developed.