Democritus University of Thrace (DUTH)
Online appearance-based place recognition and mapping algorithms for autonomous robot navigation
Abstract
dc:descriptionIn the field of technology and research, autonomous robotics systems constitute one sector in the state of the art. The recent robotic systems are constructed in a way to be able to determine their pose while they navigate through the environment. This goal is achieved due to the construction of a consistent topological map while the robot crosses a route, a technique wide known as Simultaneous Localization and Mapping (SLAM). The map creation is based on the data measures which come from the robot’s onboard sensors. The increase of computation power in recent years and also the wide use of cameras lead the autonomous systems to use optical sensors, as the main incoming data source. Since the robot traverses its route into an unknown environment the danger of construction inconsistent map is very large. This is due to noisy sensor measures or possible devices’ malfunction. Those cases may introduce great risk in robotic’ autonomous missions. Difficult situations, as one mentioned before, can be fixed with the rectification of the robot’s trajectory if the traversed place can be recognized through the data of optical sensors it is equipped with. The systems which are based only on cameras for place recognition are known as appearance-based systems and the techniques which try to solve the problem of relocalization are known as loop closure detection methods. The present Ph.D. desertion focuses on the construction of a system capable to recognize a repeatedly visited place through the incoming images. The achievement of the system is based on three major units. First comes the image processing unit which is capable to provide the system with scene interpretation. Due to image description techniques local key points, which are capable to describe the observed scene, are received. Each image may contain hundreds of local key points as a result can be described as an aggregation of individual features. The map constitutes the unit which is responsible for world representation of the robot’s visited places based on the received features. At last, the belief generator unit combines incoming data with the map data as a result a decision is received if the system belongs or not to a pre-visited place and so to rectify its trajectory.
Degree
thesis:*- Grantor dc:publisher
- Democritus University of Thrace (DUTH)
- Year dc:date
- 2021
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Tsintotas, Konstantinos
- Τσιντώτας, Κωνσταντίνος
Subjects
dc:subject × 22- Αυτόνομη πλοήγηση
- Πλοήγηση ρομποτικών συστημάτων
- Οπτική πλοήγηση
- Προσαρμοστική πλοήγηση
- Αναγνώριση χώρων
- Αλγόριθμοι
- Εντοπισμός
- Αναγνώριση κλεισίματος βρόχου
- Autonomous navigation
- Mobile robot navigation
- Visual navigation
- Adaptive navigation
- Place recognition
- Algorithms
- Localization
- Loop-closure detection
- Επιστήμη Ηλεκτρολόγου Μηχανικού, Ηλεκτρονικού Μηχανικού, Μηχανικού Η/Υ
- Επιστήμες Μηχανικού και Τεχνολογία
- Ρομποτική
- Electrical Engineering, Electronic Engineering, Information Engineering
- Engineering and Technology
- Robotics
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/50432