Aerospace in the century of powered flight is a field of robots and algorithms, fueled as much by nimble startups as mass production. Lockheed Martin, a defense giant that thrived in the first century of powered flight, is looking to adapt to this new bleeding edge, while maintaining its considerable industrial presence. To that end, the company announced Sept. 5 a partnership with the Drone Racing League to test drone-piloting AI against the best human drone pilots in the world in what it is calling the AlphaPilot Challenge.
By requiring all competitors fly the same model drone, the Drone Racing League (DRL) is more akin to NASCAR than Formula One. The challenge is about correctly piloting the same stock vehicle through a three-dimensional obstacle course, not so much about building an aerodynamically faster machine.
Entrants in the competition will work with the DRL stock drones and an NVIDIA GPU, ensuring that the sensors and hardware for every team are the same. Like how stock car races test the skill of drivers, competition is tightly focused on teams building the most effective AI to load into that same, standard package. The races will start in 2019 and conclude the next year.
“In 2020, there will be a grand finale challenge to see which machine is going to be faster than the world’s best, fastest human pilot,” said Robie Samanta Roy, vice president of tech strategy and innovation at Lockheed Martin.
There’s $2 million available in prizes for the competing teams. On top of that, there’s an additional $250,000 award for the first team whose AI outflies that fastest human pilot. This would be the possible ‘Deep Blue’ moment, a callback to the 1997 chess match where IBM’s Deep Blue computer became the first machine to beat a reigning human world champion.
“That Deep Blue moment will come,” says Samanta Roy. “Maybe it’s my hand-eye coordination — these drones are like a blur when they’re going through these obstacles, right? The Deep Blue moment will be can a machine go as fast as a human pilot and beat them in a highly complex environment.”
Flying autonomously and quickly through complex indoor environments is a challenge not just for drone racing, but also for future military applications, too. For the military industry, this has been tackled head-on in DARPA’s “Fast Lightweight Autonomy” program. While the lessons of AlphaPilot may play directly into those same tasks, Samanta Roy instead sees the project as about creating a new pipeline into the industry and the company itself.
“Really, at the end of the day, this is about talent acquisition,” says Samanta Roy. “If they win, we want them to come and work for the world’s largest aerospace and defense company.”
Lockheed Martin is specifically reaching out to undergraduate and graduate students, drone enthusiasts, coders and other technologists. Interested teams can sign up online through Lockheed Martin. The company suggests that this is just the first of many such challenges, designed to feed the pipeline through getting people to engineer solutions to interesting problems.
“This is about perfecting the technology so that the machine is gonna be a better partner with the human,” says Samanta Roy. “Let the humans do what the humans are good at, let the machines do what machines are good at. It’s symbiotic.”