Self-Driving Car Faces Off Against Pro On Thunderhill Racetrack 151
Hugh Pickens writes "Rachel Swaby reports that a self-driving car and a seasoned race-car driver recently faced off at Northern California's three-mile Thunderhill Raceway loop. The autonomous vehicle is a creation from the Center for Automotive Research at Stanford (CARS). 'We tried to model [the self-driving car] after what we've learned from the best race-car drivers,' says Chris Gerdes (who talks more about the development of autonomous cars in this TED talk). So who won? Humans, of course. But only by a few measly seconds. 'What the human drivers do is consistently feel out the limits of the car and push it just a little bit farther,' explained Gerdes. 'When you look at what the car is capable of and what humans achieve, that gap is really actually small.' Because the self-driving car reacts to the track as if it were controlled in real time by a human, a funny thing happens to passengers along for the ride. Initially, when the car accelerates to 115 miles per hour and then brakes just in time to make it around a curve, the person riding shotgun freaks out. But a second lap looks very different. Passengers tend to relax, putting their faith in the automatically spinning wheel. 'We might have a tendency to put too much confidence in it,' cautioned Gerdes. 'Watching people experience it, they'll say, oh, that was flawless.' Gerdes reaction: 'Wait wait! This was developed by a crazy professor and graduate students!'"
BRAKE (Score:5, Informative)
brake brake brake brake brake brake
the word is brake
Re:cool (Score:5, Informative)
Try this: http://youtu.be/YxHcJTs2Sxk [youtu.be]
Re:Seconds? (Score:4, Informative)
AC is correct.
In motorsports few seconds is a very long time. The lap times are not mentioned.
The raceway in question is probably this: http://thunderhill.com/staticpages/index.php?page=TrackMap [thunderhill.com]
But which variation? Long version 2.866miles record times tend to be just over the 2minute mark for somewhat normal cars.
Short version is 1.769miles for which SCCA website is missing the record times, the medium version is 1.814miles and record times tend to be close to 1:30 mark with somewhat regular cars.
Also they don't say how good race car driver was the AI against, there is a huge variety of race car driver skill levels.
Few seconds? They are being vague, i bet it was more than just 2 seconds because they are being vague. :)
Some racing series have 3% rule to qualify, ie. within 3% of the best time, for 1:30 lap time that is 2.7seconds, in other words this AI wouldn't even qualify.
All that being said, great work! Got to start from somewhere.
In theory AI could become better than humans, but then again AI will most likely always lack intuition, so could well be that a human will always surpass AI.
Nevermind that a very highly skilled human with very high motivation can do some insane reaction and completely remove the guesswork some of the time when surpassing the limits, ie. see Ayrton Senna. For AI we'd need sensor capable of few ms polling rates with data returned, then compute all the data within few milliseconds and then some insane fast and accurate servos to achieve that level.
Few millisecond polling rate doesn't sound like much until you realize that for example USB has 90ms, PS/2 is in theory capable of 5ms, and serial port even faster, but that doesn't account for data transfer rates.
There's a reason why we cannot even build a simple ECU/EMS with standard off-the-shelf hardware: Polling rates are too slow.