The future—autonomous and otherwise—and the past
This is a Web-exclusive companion piece to "Balancing act," which we published in our March 2005 issue.
-- Test & Measurement World, 2/28/2005
Anthony Levandowski, the 2005 Test Engineer of the Year, elaborated on the future of autonomous vehicles as well as on his future and on technology’s past and future.
He cited specific benefits of two-wheeled autonomous vehicles. First is cost. “The military wants autonomous vehicles that are cheap and disposable so they can be deployed where it’s too risky to send a $20 million manned tank. Second, a two-wheeled vehicle can access many places where a larger vehicle can’t go. Third, two-wheeled vehicles have a form factor that makes them easy to carry to the deployment point-they can simply be picked up and loaded onto a trailer.
“I think that if you want to get results quickly you wouldn't want to go with two wheels, but I'm interested in doing this for the long term. So I can afford to actually think what the future will look like—as opposed to what next week will look like or next month will look like.”
Nevertheless, the two-wheeled approach presents many challenges in addition to the obvious ones of balance, navigation, and obstacle avoidance. One challenge, he said, stems from the fact that machines have no understanding of fear: “When in doubt, humans slow down, but computers are likely to do something very stupid. We try to build in the underlying behavior—a subconscious feeling—of how to maintain balance.”
That “subconscious feeling” inevitably will fail to keep the machine upright. To deal with that eventuality, Levandowski and his team have developed a motorized arm that can right the bike should it fall over. This approach acts in lieu of an automated “kickstand” that could deploy automatically as the bike slows to a near standstill. “It’s better to fall down and get back up than to slow down and risk catching on something that would really through you off” while attempting to deploy the kickstand.
Does he have the engineering-power assembled to address all such aspects of the project? “We could use more help, but we have all the bases covered. There’s no lack of knowledge on the part of the team, just lack of time, equipment, and funding. In fact there’s almost no time left, and we could spend all of it testing the vehicle and repeating the tests.”
In fact, he said, “The challenge is all about repeatability—just because it worked once the day before or once six months ago doesn’t mean it’s going to work the day of the race. There are so many variables—maybe it’s the first time you leave it outside overnight because you’ve always put it in garage and now there’s condensation inside vehicle that's shorting stuff. You can never do enough testing. And unfortunately we won’t have time to do extensive race-level testing, which is what the favored team is doing right now: counting quality miles, avoiding obstacles, going at high speeds of 20 to 30 mph.”
Despite the fact that the lack of time makes winning the $2 million prize unlikely, Levandowski does hope for a satisfactory showing-completing the race in 20 hours, for example. In fact, he said, “even going five miles would be incredibly satisfying-seeing the vehicle go is really uncanny because it does seem as if it’s alive on some level, constantly starting to fall and catching itself. People are sponsoring us not because they want to be on the winner’s podium but because they want to be part of the next generation of autonomous vehicles—that's what it’s really about.”
Discussing his past, Levandowski said he has always been interested in engineering. That makes him different from last year’s Test Engineer of the Year winner, metrology lab manager Chris Grachanen ("Calibration's standard bearer," March 2004), who had little early interest in the profession. “I love Legos,” said Levandowski. “That’s how I started out wanting to be an engineer. I love to see an idea move from thought to paper to reality—that’s a huge thrill.”
He contrasted his current approach to engineering with Grachanen’s: “Chris is a metrologist, and it’s all about precision and accuracy and following procedures and keeping good records. For us, well, we have instruments around that probably need calibration-many older than the team members—but we just don’t have the time and money for that. We just have to do whatever it takes” to get the project moving forward.
When asked what the greatest technical innovation of the past hundred years, he also differed from Grachanen, who, an employee of Hewlett-Packard, reported the perhaps politically wise choice of the personal computer. According to Levandowski, “A cell phone is pretty nice, but at the same time, how can you not mention the Internet? Those were the two I was debating.
“What’s really innovative about the Internet is not so much the ability of a computer to talk to a computer over a network-that was inevitable once you have a modem.” But, he asked, what is it that makes such a network useful? “You can use it when you know where you are going, but it’s even more powerful when you don't know where you’re going-when you don’t know what you want. The best tech innovation for that is Google, which let’s you access any information quickly, increasing your brain size by a million.
“So while the cell phone is nice because its mobile, but I’ve got to give it to Google.”
He also offered some speculation on the next hundred years: “Well, now you can get the Internet over a cell phone, but the user interface is so bad no one does it. How about a bio chip inside your brain that can do an instant Google search from anywhere wirelessly?” He added that that’s probably too complicated to be realized within the next hundred years.
Meanwhile, he remains focused on autonomous vehicles and hopes to commercialize some aspects of the autonomous motorcycle technology developed for the DARPA challenge. “I have had offers” to commercialize aspects of the project, “but I have to finish this. Had we gone a couple miles last year I might have been able to detach myself but feel we have unfinished business and must close loop.”
Back to "Balancing act" in the March 2005 issue.
