See the First Fully Autonomous Tandem Drift

With the goal of advancing the potential of AI to improve safety, Stanford Engineering researchers have achieved the world’s first fully autonomous tandem drift. Both cars are driverless, and, using only AI, the lead must plan and execute its line without human input. The chase car must follow suit. Watch this video to learn more.

Transcript

00:00:02 (intriguing music) (intriguing orchestral music) (intriguing orchestral music continues) (echo booms) (echo booms) (intriguing orchestral music continues) (team speaking indistinctly over radio) (bass booms) (orchestral music) (orchestral music building up) (team speaks indistinctly over radio)

00:00:38 (buttons beep) - [Team Comms] 30, beginnings from (indistinct) (metal bangs) (engines roar) (tires screeching) (orchestral action music) (engine roars) (tires screech) (wind booms) (engines roaring) (tires screeching) (metal rings) (action music) (engines roaring) (tires screeching)

00:01:11 (intense action music) (engines roaring) (tires screeching) (engine whooshes) (engines roaring) (tires screech) - The days leading up to this experiment were, without a doubt, the most challenging I've had in my professional career. - Nothing prepared us for actually getting in the car and flipping that switch to go autonomous,

00:01:41 just the visceral feeling, the vibration. - The engines are loud, the tires are making a lot of noise. The cabin is filling up with tire smoke. Ultimately, when you're in the vehicle, you're kind of along for the ride. - We've done something that nobody else in the world has done. We've shown two vehicles tandem drifting autonomously, where the following vehicle is actually reacting in real time to the first vehicle

00:02:02 and avoiding crashing. - Drifting actually puts the car in a situation where it's very controllable. - One way that we chose to do that is using a neural network vehicle model. Rather than trying to develop with physics and by hand-tuning parameters, we decide to learn through data. - One of the benefits of artificial intelligence is allows us to build a system

00:02:21 that's actually improving over time, much like an expert driver would. - We've gone from being somewhat skeptical of using AI in these safety-critical situations to really coming to believe that, for this project, we can't do this without it. (cool action music) After a number of years of working on the test track, developing these technologies, we finally got to the point

00:02:42 where we were confident in putting two cars together in a true tandem autonomous drift. - Tandem drifting is perhaps the ultimate expression of car control capability. - The cars have to be constantly reacting, constantly making small changes to the steering, the throttle, and the brakes. - Just a slight wrong move can cause a vehicle to spin out. - The way you control the car while drifting is the same as the way you would control it

00:03:05 while it's on ice or snow. That's really where the average driver can benefit from a system that's assisting them. - We'd hope that if we can do it here in tandem drifting, then we can do it safely in other scenarios. - The things that we've been developing here can one day make the roads much safer for everybody. - Doing things that have never been done before show you what's possible. They give you new ideas that you can use

00:03:29 to change the future. (intense action music)