VSLAM analysis using various ORBSLAM parameter settings

Hussin, Hishamudin (2022) VSLAM analysis using various ORBSLAM parameter settings. Project Report. Universiti Teknikal Malaysia Melaka, Melaka, Malaysia. (Submitted)

Preview

Text (Full Text) VSLAM analysis using various ORBSLAM parameter settings.pdf - Submitted Version Download (5MB) | Preview

Abstract

Navigation system is one of most crucial components in the development of robot applications. A precise navigation system helps the robot sense and map the environment to move around more efficiently. There are many robot navigation systems that have been developed to make autonomous robot such SLAM and Neural Network System. It is expected that autonomous robot can help many industries due to it capabilities that can save many human powers and save a lot of cost in the manufacturing. However, not all the developed navigation systems today provide details and precise information about the surroundings. SLAM or simultaneous localization and mapping system, is a system that determines the orientation and position of a robot by creating a detailed map of the environment while simultaneously tracking where the robot is within that environment. Meanwhile, VSLAM is a SLAM system that uses a camera to supply all the essential information of the surroundings to be processed by the system. Many SLAM projects that are being actively used in the industries use the LiDAR sensor, which is very expensive and has limited information collected. Camera has been used to overcome those LiDAR problems due to lower cost and more reliability compared to LiDAR. Replacing the LiDAR with the camera will bring much more detail and dynamic in the robot navigation system. ORBSLAM algorithm are the main focus of this project and a few parameters settings in the algorithm were tested to get the most compatible setting with the surrounding. All the parameter performance and accuracy are analyzed using the ATE and RPE evaluation method to see the most optimize parameter settings before being used in the real-time surroundings.

Actions (login required)

View Item