Aristotle University Of Thessaloniki (AUTH)
Μέθοδοι εκτίμησης στάσης αισθητήρα LIDAR δισδιάστατων μετρήσεων μέσω ευθυγράμμισης πραγματικών με εικονικές σαρώσεις
Abstract
dc:descriptionDuring the last decade mobile robotics has been on the rise in terms of its range of use and content of applications. Autonomous ground and aerial robots are used in warehouses for inventorying and localising products, in hospitals for decontamination purposes, in agriculture for monitoring crops and process automation, and for transporting goods or people. In all cases the self-knowledge of a vehicle’s pose in space, i.e. its position and orientation, is essential for its autonomous motion and the satisfaction of objectives of its mission. In indoor areas, where pose information transmitted via satellite is denied, or contexts where indoor measuring infrastructure is absent, the only alternative is the estimation of the pose vector, and the estimation process is undertaken by the vehicle itself, based on the sensors it is equiped with. The subject of this dissertation is autonomous ground mobile robotics in the context of the absence of pose measurements. Its main content is conserned with the estimation of the pose of a mobile base on the basis of two dimensional measurements derived from a lidar range sensor and the reduction of the estimation error, with the pose estimate being derived from traditional methods of the literature. Specifically we explore ways in which matching measurements from the physical sensor against virtual scans—computed either from the map of the environment or from a second measurement—can be used to satisfy the two above-stated objectives, with performance superior to that of established and state-of-the-art methods of the literature. We first design a methodology for evaluating autonomous navigation algorithms, and through its implementation we observe the existence of pose estimate errors. We then investigate ways of reducing these errors with methods intenal and external to the mechanisms of a particle filter. One of these external ways is prosthetic scan–to–map-scan matching. Discovering that scan-matching is performed in the literature in a manner that is not robust to measurement noise, the magnitude of position estimate errors, and tuning parameters, we propose scan–to–map-scanmatching methods that replace the relevant pathogenic state-of-the-art mechanisms with mechanisms based on the Fourier transform and its properties. At the application level we design systems that given the map of the mobile base environment aim at solving the global localisation problem, and reducing the pose error during pose tracking. Finally, we generalize our approach in the absence of a map in order to generate odometry based on a panoramic lidar’s range measurements.
Degree
thesis:*- Grantor dc:publisher
- Aristotle University Of Thessaloniki (AUTH)
- Year dc:date
- 2023
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Filotheou, Alexandros
- Φιλοθέου, Αλέξανδρος
Subjects
dc:subject × 14- Ρομποτική
- Εύρεση θέσης
- Ευθυγράμμιση σαρώσεων
- Αισθητήρας LIDAR
- Robotics
- Localisation
- Scan-matching
- LIDAR sensor
- Επιστήμη Ηλεκτρολόγου Μηχανικού, Ηλεκτρονικού Μηχανικού, Μηχανικού Η/Υ
- Επιστήμες Μηχανικού και Τεχνολογία
- Αυτοματισμοί και Συστήματα ελέγχου
- Electrical Engineering, Electronic Engineering, Information Engineering
- Engineering and Technology
- Automation and control systems
Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/56399