AbstractsComputer Science

3D Scan-based Navigation using Multi-Level Surface Maps

by Andreas Persson




Institution: Örebro University
Department:
Year: 2009
Keywords: Natural Sciences; Computer and Information Science; Computer Science; Naturvetenskap; Data- och informationsvetenskap; Datavetenskap (datalogi); Engineering and Technology; Teknik och teknologier; Natural Sciences; Computer and Information Science; Naturvetenskap; Data- och informationsvetenskap; Natural Sciences; Computer and Information Science; Software Engineering; Naturvetenskap; Data- och informationsvetenskap; Programvaruteknik; TECHNOLOGY; Information technology; Computer science; TEKNIKVETENSKAP; Informationsteknik; Datavetenskap; TECHNOLOGY; TEKNIKVETENSKAP; TECHNOLOGY; Information technology; TEKNIKVETENSKAP; Informationsteknik; TECHNOLOGY; Information technology; Computer science; Software engineering; TEKNIKVETENSKAP; Informationsteknik; Datavetenskap; Programvaruteknik; Computer Engineering; Datateknik; teknik; Technology
Record ID: 1345551
Full text PDF: http://urn.kb.se/resolve?urn=urn:nbn:se:oru:diva-11211


Abstract

The field of research connected to mobile robot navigation is much broader than the scope of this thesis. Hence in this report, the navigation topic is narrowed down to primarily concerning mapping and scan matching techniques that were used to achieve the overall task of navigation nature. Where the work presented within this report is based on an existing robot platform with technique for providing 3D point-clouds, as result of 3D scanning, and functionality for planning for and following a path. In this thesis it is presented how a scan matching algorithm is used for securing the alignment between provided succession point-clouds. Since the computational time of nearest neighbour search is a commonly discussed aspect of scan matching, suggestions about techniques for decreasing the computational time are also presented within this report. With secured alignment, the challenge was within representing provided point-clouds by a map model. Provided point-clouds were of 3D character, thus a mapping technique is presented that provides rough 3D representations of the environment. A problem that arose with a 3D map representation was that given functionality for path planning required a 2D representation. This is addressed by translating the 3D map at a specific height level into a 2D map usable for path planning, where this report suggest a novel traversability analysis approach with the use of a tree structure.