Evaluating the Usability of DronePIT: A VR Drone Pilot System with Expert Pilots

Abstract

In the last few years, industries like infrastructure, energy, and emergency services have started using drones for aerial inspections due to their low cost and size. Yet, training drone pilots remains costly, time-intensive, and limited by physical constraints. One way to solve this issue is to use Virtual Reality (VR), as an alternative for traditional courses. This thesis presents \system, a VR drone training system designed in collaboration with drone industry professionals that incorporates real-world training scenarios. The system has three phases to simulate learning, navigation, and inspection tasks. Twelve experienced drone pilots evaluated \system in a usability study. Participants interacted with \system, completed standardized usability and workload questionnaires (including the VRSQ, and NASA-TLX), and participated in post-experience interviews. Behavioural metrics such as trajectory efficiency, task completion times, and control stability to understand pilot performance were also captured.

Results indicate that participants perceived \system as highly usable, with minimal simulator sickness and low cognitive load. Participants also reported that the virtual flying experience closely matched their real-world expectations and habits. Behavioral data showed that expert pilots navigated and controlled the drone in VR in ways that mirrored their real-life techniques, supporting the idea that VR can elicit authentic motor and decision-making patterns when designed appropriately. This thesis contributes to the broader field of VR-based training by highlighting how evaluating task-specific systems with expert users can improve learning outcomes and reduce the cost and risk of VR training systems.