Institutional Repository
Technical University of Crete
Technical University of Crete
EN
|
EL
| URI: | http://purl.tuc.gr/dl/dias/00D4F34B-FC5A-4EC7-B4B0-E4C702E62C9F | ||
| Year | 2019 | ||
| Type of Item | Diploma Work | ||
| License |
|
||
| Bibliographic Citation | Nektarios Sfyris, "Hybrid visual simultaneous localization and mapping (SLAM) on the Nao Robot using ROS", Diploma Work, School of Electrical and Computer Engineering, Technical University of Crete, Chania, Greece, 2019 https://doi.org/10.26233/heallink.tuc.82763 | ||
| Appears in Collections |
Simultaneous Localization and Mapping (SLAM) is one of the fundamental problems a robot must solve in order to become truly autonomous. A variety of SLAM methods have been proposed, depending on the available robot sensors for measurements (camera, laser, infrared, lidar, sonar, GPS, compass, etc.) and the available prior knowledge about the environment being mapped and navigated, ranging from controlled environments, such as robotic warehouses, to totally unstructured and unknown terrains, such as the scene of a disaster. In this thesis, we present a Hybrid Visual SLAM approach, implemented and tested on the monocular case of the Nao humanoid robot. The proposed approach combines the benefits of both a Direct (Direct Sparse Odometry or DSO) and an Indirect (Oriented FAST and Rotated BRIEF SLAM or ORB-SLAM) visual odometry method. Specifically, the Direct module provides the total system's initialization process and local camera tracking, while the Indirect module provides relocalization, loop closing and map refinement. In addition, points of the physical three-dimensional space are selected from each module and at each camera keyframe to create the final consistent sparse point cloud map of the environment. All these tasks are executed in a parallel and multi-threaded architecture on a remote computer station, which communicates with the robot over a wired or wireless network. To increase the system's efficiency, we have also included both a geometric and a photometric calibration method to correct the camera measurements. Communication between the Direct and Indirect modules, as well as between the robot and the remote computer station, takes place within the Robot Operating System (ROS) framework, which enables for a common message transmission protocol. Last, but not least, a teleoperation node is built to simulate autonomous robot navigation during SLAM. The coupled system applied to the Nao humanoid robot is evaluated in various indoor and outdoor environments to demonstrate its robustness and real-time performance.
Copyright © DIAS 2013
