Autonomous controller for unmanned aerial vehicle

Abstract

The Unmanned Aerial Vehicle (UAV) is one of the technologies that is constantly evolving in order to be implemented for more applications. In the past, it is only used for military application for combat purposes, where the UAV will carry aircraft ordnance such as missiles for air strikes. As the technology for UAV advances and the cost of development reduces, it becomes a technology that is readily available for the common people. Now, UAV can be applied in more areas, such as photography, delivery and critical missions such as Search & Rescue. One of the techniques to further improve the performance and utility of UAV is to automize it and giving it the ability to operate on its own. While research has been done on the potential for autonomous UAV, the challenge has been to provide a stable and consistent controller for it. Due to the complexity of the hardware of UAV and the directions that it is required to operate on, it is difficult and time consuming to tune and optimize the UAV to operate perfectly under every condition. This project explores the performance of Parsimonious Automatic Controller (PAC) when implemented on an autonomous UAV and compares the result with the commonly used linear Proportional-Integral-Derivative (PID) controller. With the application of machine learning implemented into PAC, it reduces the parameters that is required while improving the performance of the controller in the process. By testing and simulating the use of this controller onto the autonomous UAV, it determines the viability of implementing the UAV for real-world missions. It also opens the possibility of implementing it on other platforms or systems and potential improving the performance of it.