Development of a Dataset to Evaluate SLAM for Fukushima Daiichi Nuclear Power Plant Decommissioning

Abstract

Simultaneous localization and mapping (SLAM) is not only a key technology for robot to move automatically, but also is a useful technology for humans to understand the state of places. Especially for a site where access is constrained by something harmful, such as Fukushima Daiichi Nuclear Power Plant (1F), understanding of the state of the site is important. In addition, mapping using remotely control robots is one of the ideal solution for investigation such sites. However, there are technical problems to solve for applying SLAM to extreme environments, for example, how to obtain landmarks under severe environmental condition. Furthermore, for extreme environments, we have very limited or no chance for testing SLAM on the actual site, and this makes it difficult to research applying SLAM. For this reason, an evaluation method without testing on the actual 1F site is needed to promote research of SLAM for 1F. This paper introduces the development of a dataset for SLAM evaluation with the mockup field instead of the actual site, specifically the dataset of the mockup field of Primary Containment Vessel (PCV) platform under dark illumination condition.