AI Driven Drone Outperforms 3 Human Champion Drone Racers

An artificial intelligence drone system known as Swift outperformed three champion human drone racers. This victory marks a significant achievement for AI, showcasing its capabilities in a real-world physical sport.

The High-Speed Contest

This high-speed drone contest took place in first-person view, where pilots operate quadcopters remotely. The drones sped through a specifically designed track at speeds surpassing 100 kilometres per hour. The track, located at Dubendorf Airport near Zurich, required intricate manoeuvres.

Swift vs Champions

The contest saw Swift take on three champion racers: Alex Vanover, who was the 2019 Drone Racing League champion, Thomas Bitmatta, the 2019 MultiGP Drone Racing champion, and Marvin Schaepper, a three-time Swiss champion. Swift managed to secure victory in 15 out of the 25 races, even recording the fastest lap time.

Difficulties Faced by Swift

While Swift’s performance was impressive, the AI drone did face some difficulties. It had trouble when conditions changed from those it had been trained for. This included changes in lighting conditions and other unexpected environmental shifts. The human pilots, on the other hand, were more adaptable in such scenarios.

The Technology Behind Swift

Swift’s AI system was a product of combined expertise from the University of Zurich and Intel. It was trained in a simulated environment, learning to fly using a method known as deep reinforcement learning.

This method involves learning from repeated practice and adjusting based on errors. Swift’s design incorporated a neural network to detect gates based on camera footage and traditional algorithms to process sensor data about speed and position. Another neural network controlled thrust and rotation speed using input from other algorithms.

This significant win for Swift is a demonstration of the advances in AI capabilities. Many believe that Swift’s design can be applied to drones for more energy efficiency.

The AI drone’s training allowed it to swiftly react in real-time, relying on data from its onboard camera. This adaptability and rapid reaction time can be beneficial for various tasks such as searching for people in emergencies or inspecting large structures.

Human Racers’ Reactions

The human racers, while gracious in their defeat, had mixed emotions about the results. Bitmatta expressed his mixed feelings, stating, “This is the start of something that could change the whole world. On the flip side, I’m a racer, I don’t want anything to be faster than me.” Schäpper added that racing against a machine felt different because the machine doesn’t tire.

Military Interest and Caution

The successful demonstration of Swift’s capabilities has caught the attention of the military, with interests in AI-powered drones. While many see the advantages of using such technology, experts caution that translating these achievements into military use might not be straightforward.

A Researcher’s Perspective

Elia Kaufmann, one of the researchers behind Swift, expressed his reservations. He mentioned that most drones in military settings are used in open battlefields for reconnaissance or against slow targets.

AI’s Future in Real-World Settings

The Swift AI drone’s success in this race marks a significant step in showcasing the abilities of AI in real-world settings. AI continues to evolve and improve, and with that, we’ll be seeing even more applications and successes in different fields.