Artificial intelligence to compete against human pilots in $2M Drone Racing League challenge

Sep. 06, 2018 07:06AM EST

The nascent sport of drone racing is already planning to have its best human pilots square up to artificial intelligence.

The face-off will begin in 2019, when a competition backed by aerospace firm Lockheed Martin will offer up a $2 million prize fund for developers who think they can code the best virtual pilot.

Read More:

Aimed to accelerate the development and testing of fully autonomous drone technologies, the Lockheed Martin AlphaPilot Innovation Challenge will begin accepting entries this November.

Selected participants will be invited to compete in a number of challenges with over $2 million in prize money up for grabs, including $250,000 for the first team that outperforms a professional Drone Racing League (DRL) pilot.

Undergraduate and graduate students, drone enthusiasts, coders and anyone else interested in drone and AI technologies are invited to enter the challenge.

The aim of the competition is to create a piece of AI software capable of flying one of DRL's standard-issue 80mph racing drones through one of the league's complex race courses. The AI has to manage this without any human supervision or having the course pre-programmed into its memory. In other words, the AI will be tasked with navigating the course as a human would, rather than following a pre-programmed flight path.

Teams will be invited to race their AI-piloted drones against each other in DRL's 2019 season, before taking on the best human pilots.

"This challenges changes the game," said DRL founder and chief executive Nicholas Horbaczewski. "How close is AI performance to the world's best human piloting? We're excited to find out next year when AlphaPilot drones compete in adrenaline-packed, futuristic drone races on complex courses in the AIRR [Artificial Intelligence Robotic Racing] Circuit."

AI racing AI is similar to the plans of Roborace, the autonomous race car series which is gearing up to become a part of the Formula E race calendar, and where teams compete using their own AI and a grid of standard-issue race cars. A Roborace prototype completed the famous Goodwood hillclimb in England this summer, navigating the one-mile course itself - albeit at a fairly low speed.