Artificial Intelligence is a disruptive technology that can profoundly change aspects of our society. Since 2012 there has been great progress in AI, thanks to the rapid developments in “deep neural networks”. The world champion in chess was beaten by IBM’s Deep Blue in 1997, humans were beaten at Jeopardy by IBM’s Walter in 2011, in Go by Google Deepmind’s AlphaGo in 2016, and in StarCraft II in 2019. In aerospace, the next achievement may consist of beating human drone race pilots with an autonomous drone.
Figure 1: The AIRR racing drone, it has an Inertial Measurement Unit (IMU), a downward pointing laser ranger, 4 high-resolution cameras, and an NVidia Xavier board: 512-Core Volta GPU with Tensor Cores, 8-Core ARM v8.2 64-Bit CPU, 8 MB L2 + 4 MB L3., 16 GB 256-Bit LPDDR4x Memory., 32 GB eMMC 5.1 Flash Storage.
In 2019, the first AI Robotic Racing competition was organized. The goal was to (1) have autonomous drones race together utilizing different approaches to AI and drone racing, and (2) have the winner compete with one of the world’s best human drone race pilots, Gab707. 424 teams from 81 countries participated in the competition. At the end, 9 teams competed against each other in 3 seasonal races and 1 world championship race. The TU Delft team became 1st in the championship but lost against the human pilot. For more details: http://mavlab.tudelft.nl/mavlab-wins-the-alpha-pilot-challenge/
Assignment specification
In the AIRR races, the drone had to pass through large gates designed by the competition organizers, the Drone Racing League and Lockheed Martin. Detecting these gates in the images from the drone’s cameras was a vital task for the competition.
