Talk:Sebastian Thrun
From Wikipedia, the free encyclopedia
 |
This article is within the scope of WikiProject Biography. For more information, visit the project page. | ||
| ??? | This article has not yet received a rating on the project's quality scale. [FAQ]
Please rate the article and, if you wish, leave comments here regarding your assessment or the strengths and weaknesses of the article. |
||
This site claims that Thrun pioneered the field of probabilistic robotics. However, the field was pioneered in the early 1990s (and even the 1980s), long before Thrun made any contributions to it. Self-aggrandizing claims like that come at the expense of others and should be removed or toned down. The field had several pioneers, but if I had to identify a single one I'd probably go for Hugh F. Durrant-Whyte, now head of the Australian Centre of Field Robotics. With Google scholar one can easily find his numerous early publications on probabilistic methods for robotics.
Response by an independent observer: Interestingly, the history of probabilistic robotics is much older, as many of the early robots relied on Kalman filters, invented in the late 1950s (Swerline/Kalman). Peter Cheeseman brought probabilistic filters into robotics by inventing the Kalman filter method for Simultaneous localization and mapping (SLAM), which was soon further developed by important researchers like Raja Chatila, Hugh Durrant-Whyte, and many of Durrant-Whyte's graudate students. It is a field with many pioneers, like John Leonard, Mark Paskin, Dieter Fox, and others. The book Probabilistic Robotics has many of the original references.
Contents |
[edit] POV: Wired Magazine's Local Patriotism & Post DARPA Race Hype
Yet another article linking to a questionable POV article by WIRED, a magazine based in close proximity to Stanford University and known as some sort of Stanford megaphone. Before the 2005 DARPA Grand Challenge, Sebastian Thrun and many others predicted that this time with very high probability there would be a winner, and in fact four separate teams managed to finish the course in time, which suggests that the task was not that hard. Previous autonomous cars actually drove further and faster, at least on streets. But after the somewhat lucky victory of Stanford's Stanley, facilitated by a technical glitch of CMU's Sandstorm, Thrun suddenly changed his mind and fueled the hype by claiming: "The impossible has been achieved!" According to the WIRED article, he later publicly compared himself to Charles Lindbergh - LOL - I felt reminded of a chicken comparing itself to an eagle. As if that were not enough, soon afterwards the Stanford AI lab home page also claimed that Stanford's Volkswagen is "the best robot ever", citing another ridiculous article of WIRED. Few if any unbiased roboticists outside the San Franciso area would agree - what kind of "experts" did WIRED poll? Maybe cartoonists, since Number 2 of the their top 10 list is a fictional Japanese comic strip robot. The other cars that finished shortly after the lucky winner are not even mentioned... Actually I don't think any mere car would rank anywhere near the top 5 of a serious robot list, which would be dominated by Japanese robots - real ones, not fictional ones - since Japan dominates research in this field and has 40 percent of the world's robots, including many of the most sophisticated and famous ones. I am not sure what's the Wikipedia policy on this, but I feel it would be better to remove the WIRED link and similar POVs. De-Hyping Stan 17:46, 10 April 2006 (UTC)
Personally, as the head of one of the other (losing) teams, I'm impressed with the Stanford Grand Challenge entry. Their vision system actually worked, which is very impressive if you understand what it was doing. I didn't think they could make that approach work. --John Nagle 07:03, 11 April 2006 (UTC)
Your posting bespeaks the noble attitude of a gracious loser. And I totally agree, one should point out impressive aspects of the software, if there are any. But that's quite different from hyping things out of proportion :--) De-Hyping Stan 17:18, 12 April 2006 (UTC)
Just so that we're clear on his "comparison" of himself to Charles Lindbergh, here's the actual quote from the WIRED article: "Some people refer to us as the Wright brothers," he says, holding up his champagne. "But I prefer to think of us as Charles Lindbergh, because he was better-looking." TNeloms 06:04, 24 October 2006 (UTC)
Also, it was apparently Tony Tether (DARPA director) who compared them to the Wright brothers in the first place. TNeloms 06:08, 24 October 2006 (UTC)
The funny thing is that Tony Tether (the guy with the inappropriate comparison to the Wright brothers) is also ex-Stanford. To summarize: Tether is from Stanford, the Stanley team is from Stanford, Wired magazine is next to Stanford ... it does look like Stanford people hyping each other. Truecobb 21:14, 12 April 2007 (UTC)
[edit] SLAM link
Can someone please clairify what the SLAM link should point to? It is just going to the disambiguation page now. --Dan 14:50, 17 April 2006 (UTC)
[edit] We all won
I really like the chicken versus eagle comparison (which are much more elaborate in the discussion section of the article on Stanley). Coincidentally, Stanley was originally called "roadrunner," since it is much more of a chicken than an eagle. For one, it doesn't fly, it stays on the ground. Which is much harder. But more importantly, the car behaves much more like a chicken than an eagle. It is quite afraid of anything. Not just rocks, hay bales, tank traps, no. Tumbleweeds, birds, large insects - all those make Stanley swerve like a frightened chicken (as do occasional computer hangups; see our forthcoming report on what made Stanley drive on the berm at Mile 22.4).
What's more, chicken have been infinitely more successful in the recent history than eagles. Their number has grown tremendously in the 20th Century, whereas eagles are now nearly extinct. They are also more successful in the movies - or have you heard of "Eagle Little?" Or "Eagle Run?" Who wouldn't want to be a chicken these days! It's the way to go.
More seriously, I totally stand behind the comments I made in the NOVA documentary: we all won. The successful outcome of the Grand Challenge has been a victory for all of us, for every team. There have been numerous amazing inventions not just in the teams who finished, but also those whose car didn't make it. An eleven-minute difference is really quite insignificant, and others like Team Ensco and Team DAD were doing remarkably well before they ran into trouble. And the Red Team drive two robots home, not just one, which is world record.
John Nagle from Team Overbot was so kind as to point out some of Stanley's technological innovations. Let me briefly state who my true heroes were in the race. I originally had two, then three. First, Anthony Levandowsky and the Blue Team, for making a motorcycle do what it did. This was absolutely amazing. I only wish Ghostrider had made it to the race. Then the kids of Palos Verdes High School. This must have been the only team lead by a girl too young to drink. Palos Verdes High School made it to the semifinals and almost into the race, and at the site visit their car was faster than most others, including Stanley. What an achievement! And finally, the Gray Team. I feel the Gray Team deserves much more attention for what they achieved. There was a great article about them in the Wall Street Journal by Lee Gomes. The Gray Team didn't really have a decade of experience in robotics, like us at Stanford or the Red Team at Carnegie Mellon University. Many more articles should be written about their amazing success, and what they managed to achieve in just a few months' time.
There are many more heroes in this amazing race, many more untold stories. This race was a victory for the entire robotics community, We all won. Hype or not, this was a step forward.
Sebastian Thrun, April 20, 2006
[edit] The long shadow of Ernst Dickmanns
Some claim there was progress since the US teams in the earlier 2004 DARPA Grand Challenge failed. Picking the 2004 race as a straw man for a good bashing! But of course the real reference point is what Ernst Dickmanns achieved ten years earlier. In 1995 his car autonomously drove up to 158 km, nearly the same distance as the traffic-free 2005 DARPA Grand Challenge, but much faster, and in traffic. Without GPS, using only computer vision, which apparently was much more sophisticated than the vision used in the GPS-based DARPA races. I cannot see any significant technological innovations or progress here. And even the upcoming 2007 Grand Challenge looks rather tame when compared to what Dickmanns did. Truecobb 21:12, 15 July 2007 (UTC)
