Development of Simultaneous Localization and Mapping Algorithm using Optical Sensor for Multi-rotor UAV

摘要

The advancement of an autonomous Unmanned Aerial Vehicle (UAV) has permeated throughout various aspect of our lives, from military to home entertainment. Autonomous UAV has bought a new level of sensation which enables the operator of the autonomous UAV to fly the UAV to the predetermined location just by selecting the location on the map projected on the screen as this can be done by implementing the GPS module on the drone. However, a slight interference will cause a tremendous effect on the GPS signal and it is impossible to operate the autonomous UAV indoor due to the weak signal. Therefore, the objective of this research paper is to develop an alternative for UAV to detect its own location up to 25cm of accuracy without the use of GPS module in a closed environment or under the tree canopy. It is also to perform a highly accurate and precise Simultaneous Localization and Mapping (SLAM) algorithm on a low-power processing unit yet robust for UAV to map and navigate. In order to have further understand on how the SLAM algorithm works, the offline simulation was carried out on a ground computer. On top of that, a monocular offline data was downloaded and simulation of monocular SLAM was carried out with the data. Once the offline data simulation was completed, Robot Operating System (ROS) was then installed in the ground computer to perform real time monocular SLAM using a webcam. From the real time monocular SLAM, the webcam was used to capture the images and pinpoint the feature points of each image. This process will slowly generating a 3D map by each key frame. The real time SLAM was then performed with a low-powered processing unit on top of the UAV for mapping and navigation. In a nutshell, the expected outcome of this research paper is to develop a low-powered yet robust processing unit for SLAM algorithm in autonomous UAV to determine its own location up to the accuracy of 25cm for navigation and mapping purpose.