It is the first time that an autonomous mobile robot has beaten human champions in a real physical sport. The paper by Davide Scaramuzza and his team was published by «Nature». Congratulations!
The AI drone «Swift», designed by the Robotics and Perception Group at the University of Zurich, combines deep reinforcement learning in simulation with data collected in the physical world.
The drone races consists in piloting around a complex 3D Circuit. The challange for the drone is to fly very fast while estimating speed and location using the onboard camera. The video stream is processed by a deep neural network to detect individual racing gates. This information is then combined with inertial sensors to estimate the drone's position, orientation, and speed on the race track.
In 2016 Google DeepMind's AlphaGo won against the top champion Lee Sedol at Go, and in 1996 IBM's Deep Blue won against Gary Kasparov at chess. These competitions where AI algorithms prevailed over human champions are key milestones in the history of AI but they were limited to board games. The AI drone «Swift» won 15 of 25 races against three of the world's best racing drone pilots at speed of over 100 km/h. This result marks the first time that an AI has beaten a human champion in a real physical sport designed for and by humans. Such as, this result is a milestone for mobile robotics, machine intelligence, and beyond.