Artificial intelligence (AI) has made significant progress in several areas and has indeed outperformed humans in specific tasks and domains. In recent years, AI has demonstrated its prowess at tasks by making effective use of its strengths in extensive data processing, pattern recognition, optimization, and handling repetitive calculations. As a result, it has achieved a remarkable level of task performance in a wide range of fields.
Continuing the trend of surpassing human performance in several domains, an AI-powered drone recently beat three world champion human drone pilots in a high-speed race competition.
According to The guard, Developed by researchers at the University of Zurich, the Swift AI won 15 of 25 races against world champions and clocked the fastest lap on a course where drones reach speeds of 80 km/h and endure accelerations of up to 5 g, enough to ensure that many people black out.
“Our result marks the first time an AI-powered robot beats a human champion in a real physical sport designed for and by humans,” said Elia Kaufmann, a researcher who helped develop Swift.
First-person view In drone racing, you fly a drone around a track littered with fences that you must pass through neatly to avoid a crash. The pilots see the course through a video feed from a camera mounted on the drone.
Write in diary NatureKaufmann and his colleagues describe a series of head-to-head races between Swift and three champion drone racers, Thomas Bitmatta, Marvin Schapper and Alex Vanover. Before the competition, the human pilots had a week to practice on the track, while Swift trained in a simulated environment with a virtual replica of the track.
Swift used a technique called Deep Reinforcement Learning to find the optimal commands to rampage through the circuit. Because the method is based on trial and error, the drone crashed hundreds of times during training, but because it was a simulation, the researchers could simply restart the process.
“To ensure that the consequences of actions in the simulator were as close as possible to those in the real world, we designed a method to optimize the simulator with real data,” said first author Elia Kaufmann.
Click for more trending news