For the first time, an AI drone pilot has beaten a human world champion.

There’s a decent chance you didn’t know drone racing was even a thing.

The sport has taken off over the past decade as drones have gotten faster, lighter, and nimbler.

Competitors race against each other through an obstacle course. The fastest drones can go from 0 to 120 mph in one second.

Here’s a clip of the race between the AI and a human…

This clip isn’t sped up. The drones are moving so fast that AI developers struggled for years to build a system that could compete at these speeds.

The study published in the scientific journal Nature outlines how programmers built Swift, the AI system to tackle the flight challenges.

The implications of this go well beyond racing. AI-piloted drones can help with search and rescue, crop and structural inspections, and even home delivery.

New Drone Racing Champ

Swift uses two sensors to navigate these courses – a camera and something called an inertia sensor.

The camera is surprisingly low-tech. It still suffers from motion blur. But even amid the blur, the AI can see course markers.

The inertia sensor provides the drone’s onboard computer system with essential data about the drone’s motion and orientation in space.

Swift competed against three champion human drone pilots…

  • Alex Vanover, the 2019 Drone Racing League world champion

  • Thomas Bitmatta, two-time MultiGP International Open World Cup Champion

  • Marvin Schaepper, three-time Swiss national champion

The AI simulates 550 hours of flying the course. It also completes practice runs to validate its path.

To make the competition fair, the human pilots had a week to familiarize themselves with the course.

Swift won 15 out of 25 races. It also won the majority of races against all competitors and set the fastest overall lap time.

Now, I doubt you care too much about drone racing as a sport. But this has important implications off the race course.

These drones run on batteries. So they’ve limited flight time.

AI technology allows them to make the most of their battery life. They’ll be able to pick the best route for whatever their mission might be… and optimize the battery life.

And that helps drones carry out a range of other tasks.

Range of Applications

Search and rescue teams already use drones in their missions.

A drone helped to spot and rescue a lost hiker in Utah’s Snow Canyon Park. In the aftermath of Hurricane Harvey, drones were deployed in Houston to help find people stranded in flooded areas. And during the 2018 California wildfires, drones were used to find missing people at night.

For the most part, these drones require a human pilot. That means search and rescue teams are limited by the number of trained drone pilots they can get.

But an AI-piloted drone with thermal imaging would be able to rapidly scan areas for missing people. And a swarm of drones could rapidly cover a larger area. This would up the chances of a successful rescue.

That’s enough to get me excited about this technology. But there are also commercial applications.

Farmers are using drones to cut the number of herbicides and pesticides they use.

With specialized sensors, pilots can spot crops that need additional treatments. Known as precision agriculture, AI will take this to the next level.

This will save farmers money on treatments… and reduce the amount of chemicals that end up in our food.

But the biggest commercial use is already underway…

Amazon is rolling out delivery drones at its one-day fulfillment centers.

For example, the company has already delivered hundreds of items via drone at College Station, Texas since 2022.

Amazon plans to deliver millions of packages in the coming years via drones.

As AI technology improves, these drones will be able to make more deliveries in a day and spend less time on charging stations.

You need to realize that this is much more than just a computer winning a drone race. It shows how AI can enable automation. The real-world applications will only build on these events.

It reminds me of when DeepMind’s AI, AlphaGo, beat the world champion in 2016. And when IBM’s Watson beat two of the best Jeopardy! players in 2011.

They’re big tech breakthroughs that get their start with competitive games… but are preludes to bigger commercial innovations.

AI-enabled automation is going to allow businesses to do more with less.

Regards,

Colin Tedards
Editor, The Bleeding Edge