A look at high-speed autonomous driving
June 8, 2009 | Length: 00:05:18
Volkswagen's Marcial Hernandez and Sun's Greg Bollella detail Project Bixby, an Audi TTS programmed by Volkswagen and using a Java runtime environment. The vehicle will then be raced on a Rally course against other automated vehicles.
Related Videos
In
I doubt luxury car makers or sports car makers
RE: A look at high-speed autonomous driving
I doubt luxury car makers or sports car makers
Transcript
music
>> You know, over the last decade we've been showing real-time Java going to more and more demanding environments, culminating in the industrial robots and industrial controllers we've had on stage here. But they've been sort of removed from the everyday experience. And that's why I was really, really excited about Project Bixby. So we have some slides. So we're actually going to automate on Audi TTS, completely no human intervention to drive at really high speeds around a rally course, using a GPS-defined rally course. And it's gonna go fast.
>> So this is, this is, this is the Urban Grand Challenge, with the wheels unstuck from the pavement -
>> Exactly.
>> - the whole time.
>> Exactly. So in the next slide we have the teams that are participating in this. So the Volkswagen insuring research lab here in Palo Alto, they're actually building and instrumenting the Audi. Of course it's drive by wires, so that we're kind of injecting ourselves into the drive by wire system. And that's part of the deal, as well as what they're doing is developing the safety strategy, and we'll talk a little bit more about that. Remember, this car's gonna be going hundred and sixty KPH, so without a driver. So safety is crucial. At Stanford the Volkswagen automotive innovation lab is developing the control algorithms for the vehicle. So they've got the, they've got the part that they use in mat lab assumed spelling, and try to figure out you know, how do we control this vehicle, and get it to do the thing we want to do, which is follow the GPS points, and make it around the course, and not hurt the bystanders. And Sun, our part in this is we're going to provide the Solaris, and real-time Java platform on which the control algorithms will execute. And we'll be helping them and doing a few other things. So we have a video now of the GTI that's sort of the previous version, with a driver. And the driver is there because well, the top form that they're using, they don't trust. That's why they came to us.
>> Right.
car engine Right. We wanted it, one of the interesting things about Java is, is all the stuff about fault handling, fault containment, fault isolation that is just pretty unique. I mean it's something that Java developers don't usually think about, but when things go wrong, they don't go spectacularly wrong. Right? We can recover.
>> So this is typically where we end up running our vehicles as a large skid pad, with lots of runoff room. We've got them on pavement so that we know exactly how it's gonna behave, what's going to happen. And Volkswagen's been involved with autonomous and automatic vehicles for a long time, and we've got here a history of them, starting with some steering robots as far back as 2000 that were used for testing driver assistance systems, and those things, culminating on the GTI of which you saw, which was doing kind of, only was able to be beaten by Formula One drivers and the like. And at the same time we've got all of the work that we've been doing on the Grand Challenge and the Urban Challenge, where we partnered with Stanford to build those vehicles, and sort of taking those things together into Project Bixby, where we want to be doing things on dirt courses, where unexpected things are gonna happen with the trail, unexpected things are gonna happen when the vehicle spins out at a turn, and we want to be able to control it. And the big thing is we don't really want to have a driver in the car for the safety of the tester, and so we're gonna pull that out, and everything has to rely on the Java real-time platform. We want to show you guys a little bit of a sneak preview of what we'll have in October on the next video. These are the algorithms that are being developed by the Dynamics Design Lab at Stanford, and this is their prototype vehicle. Right now we're working on taking all of this code, and beginning to run it on our Audi TTS.
>> Right, so in this platform the code's pretty fragile, has timing issues.
>> Oh yeah.
>> Yeah, yada-yada.
>> Yeah. If it goes over, if it runs a little bit slow on one side the whole vehicle will freeze, and they have to step on the brakes and try to wrestle it back into control.
>> Right.
>> Not so good.
>> Right. So your vision for this isn't, I mean this looks like pure unadulterated fun. But you can actually do something useful with this, right? I mean -
>> Exactly. So on the more useful side of things, we can use this for testing, the stability control and the ABS systems by being able to do repeatable passes of the vehicle, kind of on the edges of control, and being able to bring it back in. And we also can develop a little bit more experience to be able to eventually get to the red button that says take me home on the car.
==== Transcribed by Automatic Sync Technologies ====



