Autonomous mobile robot with ORB-SLAM 2

• Main Author: Salah Eddine, Benahmed
• Format: Thesis
• Language: English; English; English
• Published: 2021
• Subjects: TJ210.2-211.47 Mechanical devices and figures. Automata. Ingenious mechanisms. Robots (General)
• Online Access: http://eprints.uthm.edu.my/6995/1/24p%20BENAHMED%20SALAH%20EDDINE.pdf; http://eprints.uthm.edu.my/6995/2/BENAHMED%20SALAH%20EDDINE%20COPYRIGHT%20DECLARATION.pdf; http://eprints.uthm.edu.my/6995/3/BENAHMED%20SALAH%20EDDINE%20WATERMARK.pdf; http://eprints.uthm.edu.my/6995/

In this thesis we introduced ORB-SLAM2. It is critical for a mobile robot to be able to identify revisited locations or loop closures while conducting Simultaneous Localization and Mapping (SLAM) in order to be successful during autonomous navigation. When creating maps, it has been determined that one of the most difficult data association problems is loop closure. It is an effective method of eliminating mistakes and increasing the precision of the robot's localization and mapping capabilities. For the purpose of resolving the loop closure issue, the ORB-SLAM method is used, which is a feature-based simultaneous localization and mapping system that works in real time. This system incorporates loop closure and relocalization, as well as the ability to do automated initialization. The monocular cameras are used to test the algorithm's performance in order to ensure that it is working properly. An important goal of this thesis is to demonstrate the accuracy of the relocalization and loop closure processes while utilizing the ORB SLAM2 algorithm in a range of different environmental conditions. The effectiveness of relocalization and loop closure in a variety of difficult indoor settings is shown via the use of a variety of experiment. According to the results of the studies, the monocular SLAM provides an accurate outcome in the interior environment. The ORB-SLAM 2 findings show the usability of the technique for autonomous navigation and future automated vehicles equipped with a low-coast monocular camera. The ORB-SLAM 2 results also demonstrate the usability of the approach for future driverless vehicles. Keywords: ORB-SLAM2; keyframe; map points localization, relocation