On the collection of robot-pose Ground-Truth, for indoor scenarios, in the RAWSEEDS project

RAWSEEDS (Robotics Advancement through Webpublishing of Sensorial and Elaborated Extensive Data Sets) is a project funded by the EEC to produce high-quality datasets, to be used in mobile robotics benchmarking. A key issue in producing high-quality datasets is the procurement of Ground Truth, so to allow a fair comparison between different approaches. RAWSEEDS focuses on benchmarking of Simultaneous Localization and Mapping (SLAM) as a mobile robotics enabling technology, and it is vital the procurement of a reliable Ground Truth for both the maps and the robot poses, to distinguish its datasets from the ones already publicly available. In particular, for the robot pose in indoor scenarios, no external devices are available for the absolute localization of the robot and to overcome this issue we devised two solutions. On one hand, we have an approach fully independent from the robot sensors; on the other, we base on the on-board Laser Range Finders, likely more accurate, for benchmarking the scientific proposals that do not require LRF streams. The project was required by its officers and reviewers to provide a validation of the robot pose Ground Truth collection system(s), i.e., an experimental evaluation of their performance; the results of such activity are presented in this paper.