In Praise of Lights and Clockwork

: A Simple Approach to Drone Autonomy

Abstract

This thesis proposes a new approach to drone automation that aims to improve safety and efficiency during inspections in complex and congested environments. Current methods have been proven effective; however, they often require a large crew including marshals, spotters, a ground station operator (GSO) and a pilot with road closures necessary in some cases to achieve acceptable safety levels. While drone autonomy research often focuses on autonomous navigation and dynamic flight planning with the aim of reducing or, in some cases, eliminating the role of the human pilot, in many civil applications such as inspection, the environment is well-known, but the possibility of map or mission error or equipment malfunction cannot be excluded.

In challenging settings, such as bridge inspections, there are different factors to consider, such as weather conditions, interactions with pedestrians and other road users, and safety monitoring the system, all of which can make the process more complicated. Autonomous systems struggle to reliably monitor all these aspects due to the complexities involved in the learning process, and they are not easily verifiable. Conversely, human operators can make more dependable judgments, such as identifying pedestrians and predicting their intentions.

This research tackles these challenges by developing a predictable and verifiable automation system that integrates seamlessly with human pilots. The proposed Ground Control Automation (GCA) system assumes responsibilities traditionally held by GSO. It employs a hazard-aware planning approach, utilizing existing maps to designate the safest rally points onsite. The GCA interface is designed to reduce pilot workload in monitoring and mission execution, thereby freeing capacity for occasional critical judgments. Although the workload is reduced, there are provisions in place for pilot engagement with the system to ensure situation awareness (SA), which is crucial for flight safety. The overall SA is managed between the pilot and the GCA, leveraging their respective strengths: repetitive tasks for the GCA and judgment-based tasks for the pilot. Ultimately, the thesis presents a formal safety analysis using inference-based safety tools to prove acceptable safety levels of using the GCA.

Date of Award	10 Dec 2024
Original language	English
Awarding Institution	University of Bristol
Supervisor	Arthur G Richards (Supervisor) & Tom S Richardson (Supervisor)