With the increase of autonomy in aircraft systems over the past decade, the question of whether autonomous agents provide better performance when compared to human operators is of utmost importance. Current research using machine learning for obstacle avoidance in small Unmanned Aircraft Systems has gained extensive attention in recent years, but who is faster, the human or the machine? This study leverages programmable coding languages using Unmanned Aircraft Systems (UAS) to navigate a fabricated obstacle course environment. By applying various methods of intelligent agent training, human pilots were compared to the performance of the automated system. The Artificial Intelligence (AI) training ensured that the automated flight path and speed for the Unmanned Aircraft System were optimized for the course environment. We collected a dataset from a sample size of 12 participants, including zero knowledge of flight training, to experienced pilots. The findings of this work aimed to show the performance comparison, including Key System Attributes and Key Performance Parameters, by considering drone speed, pilot skill level, and course crashes between man and machine to determine which is a safer and faster solution.
Human Vs. Autonomous Agents: Drone racing and Obstacle Avoidance
2022-09-18
1862829 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Are We Ready for Autonomous Drone Racing? The UZH-FPV Drone Racing Dataset
BASE | 2019
|Drone obstacle avoidance using real-time wind estimation
Europäisches Patentamt | 2021
|