Autonomous Drone Racing With The Drone Racing League
So we are here at the air race, which is artificial intelligence, robotics racing. There is code that was put onto an SD card put into the drone, and then they push go and the drones are on their own, so it is quite amazing. We'Ve got DRL pilots here, we've got all the programmers here who programmed this stuff. We got a ton of fans, we got a ton of people from the drone community and I'm excited to be here, so let's go check it out. We wanted to have autonomous drones racing from the very beginning of TRL. We always dream that we could put like an AI in among the humans competing, but we realized how far the technology was. We were on a mission to figure out how to evolve that technology quickly, so we could compete with humans, Matt Lockheed Martin. They were on the same mission, so it's a surreal meeting of the minds. We brought the technology. The event know how the media Lockheed brought their support. They brought the the sort of global presence that they have to it, and it was the perfect pairing to bring this program to life Music. We wanted to do something really visible, really fun exciting around AI and we decided to partner with the drone racing league, who was also interested in the prospects of AI, and they created a new sister league from their human. There Allianz professional racing series called air which stands for artificial intelligence, robotics racing, and this is the inaugural season of that.
Using nine teams selected to the Alpha pilot challenge. The first few levels were kind of rough I'd say because it's a bit of like a hackathon, like the teams, had no extends on experience with the drone. Really all they had was like a code that they wrote themselves. So you just turn up put it on. Then then whatever happens happens and it took them quite a bit of time to start working through it and work out the kinks and make it work better. We had one team make two gates, I think in the first level, the second level we saw one team make it all the way to the end and two level: three, where we actually saw two teams make it to the very end. So, just for everybody at home, because I don't know if people really get this, is anybody flying these things or or are they truly flying themselves? All the computers on board? The algorithm is entirely in charge of the drone. We have some well placed safety features. Of course, but all the computers on board the team's hand us the algorithm. We put it on the drone and off we go Applause, Applause, Applause, Applause, Music and just so, we understand the technology that these things are using they're, not using radar they're using vision. Sensing is that correct? Yes, so these drones are entirely dependent on the cameras in front of them, so they're doing computer vision. I like to say these drones see the world the way we do with two stereoscopic cameras: that's.
What we got that's, how we perceive that's. We make all of our decisions. The drones are the same way, so they don't have lidar and radar and other complex sensors. Those sensor packages are expensive, they're heavy they break when they crash they don't deal with vibration, very well, they're, not a great fit for drones. The future of autonomous drones, you're computer vision – and this is the stepping stone today in practice, we saw a team complete the course in ten seconds flat. Now that is getting pretty close to how fast the human pilot is able to run. That course, on the same drone, so we might see a pretty close rivalry between the AI and the human today, if not the AI surpassing the human on this course and it's actually quite nerve wracking, because so, as I watched them grow, I thought you know they. I got a Thursday, they can't finish I'm, just gon na come in and then just putts to the end and it's gon na be a walk in the park, so I'm sitting in the goggles and the thing is flying itself. So let me tell you that is very scary and you were watching first person so I'm sitting there and first of all the aircraft goes grouped up. Like oh wait, a minute I can't I could never imagine doing that and then the thing goes full tilt bah. So he goes straight for this, this pillar and then last moment the drone levels out goes pop pop and then go and then now we're through the gate like it's, not curving, it's, making almost 90 degree turn.
I could have never ever ever do that Music Laughter in history and so have you seen an improvement. This is a fourth that didn't sit in the series right. Have you seen an improvement from race one to today tremendously so at race? One we saw a couple of teams, make it to a quarter of the course route, a pretty simple course route. Today we saw multiple teams finish the course shot and finish it very quickly, nice. But at the end of the day on this one gab gab was a little worried. Didn'T look like he needed to be worried. Would you say no, he did not need to be worried. I think we've got a couple of years until I ask close to a human's capabilities. A gap is a tremendous pilot. I think he's safe for now all right, so I hope you enjoyed this. Obviously there is a lot of potential in these things. They are getting better and better with each race, but of course, cab was a little worried that he was going to get beat by the artificial intelligence. He did not. He crushed them, and that was pretty amazing, but I think that there will be a closer and closer race each time artificial intelligence gets better smarter. It learns every time I think, it's really interesting, really cool. If you like this video, please comment below give it a big thumbs up and help you hopefully you'll subscribe to ready, set your own thanks for watching and we'll, see you next time.