UCL Discovery
UCL home » Library Services » Electronic resources » UCL Discovery

Using learning from demonstration to enable automated flight control comparable with experienced human pilots

Baomar, Haitham Ahmed Omer; (2020) Using learning from demonstration to enable automated flight control comparable with experienced human pilots. Doctoral thesis (Ph.D), UCL (University College London). Green open access

[thumbnail of Thesis HB.pdf]
Preview
Text
Thesis HB.pdf

Download (21MB) | Preview

Abstract

Modern autopilots fall under the domain of Control Theory which utilizes Proportional Integral Derivative (PID) controllers that can provide relatively simple autonomous control of an aircraft such as maintaining a certain trajectory. However, PID controllers cannot cope with uncertainties due to their non-adaptive nature. In addition, modern autopilots of airliners contributed to several air catastrophes due to their robustness issues. Therefore, the aviation industry is seeking solutions that would enhance safety. A potential solution to achieve this is to develop intelligent autopilots that can learn how to pilot aircraft in a manner comparable with experienced human pilots. This work proposes the Intelligent Autopilot System (IAS) which provides a comprehensive level of autonomy and intelligent control to the aviation industry. The IAS learns piloting skills by observing experienced teachers while they provide demonstrations in simulation. A robust Learning from Demonstration approach is proposed which uses human pilots to demonstrate the task to be learned in a flight simulator while training datasets are captured. The datasets are then used by Artificial Neural Networks (ANNs) to generate control models automatically. The control models imitate the skills of the experienced pilots when performing the different piloting tasks while handling flight uncertainties such as severe weather conditions and emergency situations. Experiments show that the IAS performs learned skills and tasks with high accuracy even after being presented with limited examples which are suitable for the proposed approach that relies on many single-hidden-layer ANNs instead of one or few large deep ANNs which produce a black-box that cannot be explained to the aviation regulators. The results demonstrate that the IAS is capable of imitating low-level sub-cognitive skills such as rapid and continuous stabilization attempts in stormy weather conditions, and high-level strategic skills such as the sequence of sub-tasks necessary to takeoff, land, and handle emergencies.

Type: Thesis (Doctoral)
Qualification: Ph.D
Title: Using learning from demonstration to enable automated flight control comparable with experienced human pilots
Event: UCL (University College London)
Open access status: An open access version is available from UCL Discovery
Language: English
Additional information: Copyright © The Author 2020. Original content in this thesis is licensed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0) Licence (https://creativecommons.org/licenses/by-nc-nd/4.0/). Any third-party copyright material present remains the property of its respective owner(s) and is licensed under its existing terms. Access may initially be restricted at the author’s request.
UCL classification: UCL
UCL > Provost and Vice Provost Offices > UCL BEAMS
UCL > Provost and Vice Provost Offices > UCL BEAMS > Faculty of Engineering Science
UCL > Provost and Vice Provost Offices > UCL BEAMS > Faculty of Engineering Science > Dept of Computer Science
URI: https://discovery.ucl.ac.uk/id/eprint/10108999
Downloads since deposit
140Downloads
Download activity - last month
Download activity - last 12 months
Downloads by country - last 12 months

Archive Staff Only

View Item View Item