On a visit to the American Museum of Natural History in New York three years ago, Dennis Hong was captivated not by the giant blue whale, or the Tyrannosaurus rex skeleton, but by the ankle bone of a timid prehistoric deer.
The “double pulley” ankle gave the creature more bounce in its stride, an evolutionary advantage that enables today’s suburban deer to bound gracefully over vegetable garden fences.
DARwIn-1: The Dynamic Anthropomorphic Robot with Intelligence, was created in 2004 at Virginia Tech.
DARwIn-3: Had a special camera system, but the camera failed, and the DARwIn-3 robots froze on the RoboCup field in 2009.
DARwIn-4: Was entered in the 2010 RoboCup in Singapore and helped bring four victories.
DARwIn-OP: also designed by Dennis Hong and his graduate students, competed in RoboCup 2011 in Istanbul.
MARS: A hexapedal robot, the Multi Appendage Robotic System has the ability to walk over uneven surfaces.
CHARLI: With $20,000 in seed money, CHARLI (Cognitive Humanoid Autonomous Robot with Learning Intelligence) was built in 2008 by Hong and his students. CHARLI-2 is lighter and stronger.
Hong took out his iPhone and snapped a picture of the diagram. He thought the concept might work nicely on his next robot.
Another time, Hong watched a mother braid her daughterÂ’s hair and wondered whether a robot could move in the same way. That led to the development of a three-legged walking machine that maneuvers about with the same braiding motion.
A Chinese water-tube toy inspired Hong to create a species of robots that move by continuously turning themselves inside out. A cowboyÂ’s lasso spurred him to build a robot that wraps around surfaces like rope.
And that prehistoric deerÂ’s ankle? It became a knee bone for CHARLI, Cognitive Humanoid Autonomous Robot with Learning Intelligence, AmericaÂ’s first full-size, two-legged, walking humanoid robot.
Hong, 40, the son of a famed Korean aeronautical engineer, is the Leonardo da Vinci of robots. Leonardo saw birds in flight and imagined a human flying machine. He studied human anatomy and in 1495 sketched what is considered the worldÂ’s first robot.
Like the artist, Hong innovates by connecting things that most people might see as completely unrelated. His visions of pulleys and gears spring to life in a workshop in the basement of the mechanical engineering building at Virginia Tech in Blacksburg.
ThatÂ’s where Hong and a tight posse of 18 engineering students operate the Robotics and Mechanisms Laboratory (RoMeLa), a relatively new and audacious entrant to the robotics field.
Virginia TechÂ’s engineering school ranks 24th in the nation, according to the latest graduate rankings by U.S. News & World Report. Stanford, MIT, Carnegie Mellon and Penn all field bigger, older, better-funded robotics programs.
But none of them has Hong.
He arrived at Virginia Tech in 2003, still in his early 30s. There, working with a small team of bleary-eyed graduate students and a shoestring budget, Hong built several of the most compelling designs to emerge in American robotics.
Hong and his students “are really trying to change the world,” said Daniel Lee, a robotics researcher at Penn who collaborates with Hong.
HongÂ’s dream, though, has always been to win RoboCup, a little-known international competition that is one of the premier academic events in robotics.
RoboCup is an annual soccer tournament for robots. Designing a robot that can find and kick a soccer ball is termed the ultimate challenge in robot design; not long ago, no humanoid robot on Earth could do it. The goal is to field a robot team by 2050 that can beat the human World Cup champions.
Victory at RoboCup would place Virginia Tech and the United States at the symbolic forefront of humanoid robotics, a field that is already adding anthropomorphic machines to battlefields and hospitals and may one day put them in American homes.
Hong first glimpsed the coveted Louis Vuitton Humanoid Cup in summer 2007, during a disastrous trip to Atlanta for that yearÂ’s RoboCup. The trophyÂ’s visit to America was brief; it was headed back to Japan, where it had sat for five consecutive years.
“I looked at that trophy. I said, ‘This is the most beautiful thing I’ve ever seen, except my wife,’ “ Hong said.
As they gazed at the cup, Hong promised his students, “We’re going to bring it home.”
Czech artist Josef Capek coined the term “robot” in the 1920s; his brother Karel later popularized it in a production of “Rossum’s Universal Robots,” with human actors playing the robot parts. The first industrial robot appeared by the 1960s, and factory robots were commonplace by the 1980s.
But it is the idea of a robot based on the human body that has always captivated Hong and the public. The first humanoid robot, called Wabot-1, was built at TokyoÂ’s Waseda University in 1973.
Humanoid robots have progressed much more quickly on celluloid than in reality, particularly in America; Rosie, the robot maid of television’s “The Jetsons,” remains a distant dream. Here, most of the money -- and, thus, the interest -- lies in military applications such as disarming bombs and conducting surveillance.
But Korea and Japan have made producing humanoid robots a national priority -- the Honda motor companyÂ’s ASIMO can run, dance and conduct symphony orchestras.
The difference in priorities is partly rooted in popular culture. In America, robots are often cast as villains, while Asian countries tend to view them as heroes.
“If you ask the typical American how they view humanoid robots, their first reaction is to think of Hollywood films like ‘Terminator,’ “ said Lee, who would collaborate with Hong on RoboCup. “Whereas, if you go to Korea or Japan, their concept of robots is more like ‘Astroboy.’ “
A 2006 report by the World Technology Evaluation Center surveyed the robotics field and concluded that the United States was “in danger of losing its leading position” in what few areas it led.
Willow Garage, a Silicon Valley company, tops the field of “personal robotics” in this country and has developed a robot that can fold shirts, bake cookies and fetch drinks. Eleven PR2s are on loan to universities and research institutions. A private citizen can buy one for $400,000.
“Where the difficult part comes,” said CEO Steve Cousins, “is where you ask, ‘Are you really willing to pay $400,000 for a robot to get you beer?’ “
At the low end, Willow Garage offers TurtleBot. Built of cheap, off-the-shelf parts, with a sensor taken from the Xbox Kinect, TurtleBot looks like a cross between a foot stool and a scale, and it costs $1,200. But it doesnÂ’t do very much.
Robot parts are expensive and unreliable. Sensors for robot “eyes” cost thousands of dollars, driving up the cost of even simple gadgets. The typical robot battery cell of today loses power in 20 minutes, said Paul D’Angio, 24, a RoMeLa researcher, “and if I stabbed it with a pencil, it would explode.”
But entrepreneurs see massive potential for humanoid robots, once costs come down. At the right price, robots could spare humans from tasks that are dirty, distant, dangerous or dull.
One recent fall day, Hong and his students were preparing CHARLI for a planned meeting in South Korea with President Lee Myung-bak. (The meeting was later canceled.)
Dressed in his customary khakis and a black T-shirt, Hong walked into RoMeLa’s basement workshop at Tech to check on Jeakweon Han, 36, a long-haired, Korean-born doctoral student and CHARLI’s lead designer. Han asked, “Professor, do you want to shake hands?”
CHARLI had become a celebrity worldwide. Hong, too, was a rising star in Korea; a new robotics magazine seems to have been created with the sole purpose of putting him on its cover.
Hong wanted CHARLI to shake the presidentÂ’s hand, but he also wanted to show off a bit. Asian robots can shake hands, but in a rigid, robotic sort of way. CHARLI has a more supple, human grip.
CHARLI lifted his arm, extended his hand to Hong and shook with a surprisingly fluid motion. Hong seemed satisfied. He paused to think, then told Han, “It’s important for the robot to look at the person’s eye when he does it.”
Around Hong in the windowless concrete-block bunker, graduate students worked at desks cluttered with robot parts, along with assorted bicycle helmets, peanut butter jars and ramen noodle packets, coffee packets and energy drinks.
A side room holds a Tormach CNC, a milling machine that cuts sheets of metal into robot parts. It belongs not to the lab but to Derek Lahr, 29, a RoMeLa researcher who asked for it as a graduation present instead of a car.
As inspiration to attain his dream, Hong built a replica of the RoboCup soccer field near the basement lab; itÂ’s known as RoMeLa II.
Hong was given first choice of lab space in the not-yet-built $100 million engineering building. Now, Hong and his students are rock stars in Blacksburg, and RoMeLa has become a mandatory stop on campus tours.
ItÂ’s open almost around the clock.
One recent morning, students walked into the lab to find Han asleep at the milling machine. “There have been months on end where nobody really goes home,” said Bryce Lee, 23, a second-year grad student from Aiea, Hawaii.
Students gather Thursday nights in the lab to watch -- and heckle -- bad action movies; if they watched good movies, they might stop working. Every now and then, they take an outing to the shooting range. (The physics involved in firing a gun are “an engineer’s dream,” one student explained.)
The students take their cues on work ethic from Hong.
Before the birth of Ethan, his 2-year-old son, Hong was known for working the same punishing hours as his most obsessive students. Parenthood has slowed him, a bit, but even now he doesnÂ’t sleep much. He seems to prize slumber mostly for the inspiration it brings. He dreams in gears and joints. He keeps a notebook, a pen and an LED light beside his bed.
Many of the ideas are worthless. The good ones go into a document file, and when funding agencies such as the National Science Foundation call for grant proposals, Hong goes to his file for fodder.
In his creative process, Hong more closely resembles Keith Richards than anyone else in the Virginia Tech College of Engineering or, for that matter, the Korean diaspora to which he belongs. “Students from Korea, they’re good at math,” Hong said. “But they lack for creativity. It’s very difficult to get them to think outside the box.”
One day, Hong began one of his regular brainstorming sessions in a classroom.
How would you supply power to a robot that must operate outdoors for extended spells? he asked.
Hands shot up. Some ideas were obvious: Solar energy. A really big battery pack. Others were novel but impractical: A nuclear reactor. Program the robot to build its own battery.
One by one, Hong wrote the concepts on the board, then erased them as the group dismissed them.
Finally, someone stumbled upon a solution that was truly odd, yet oddly inspired: The robot could generate chemical energy by eating bugs. Here, finally, was something worthy of HongÂ’s database of clever ideas.
Hong has someone trained in every aspect of robot design, a dauntingly broad endeavor that spans the fields of computer programming, locomotion, vision and machining. This depth brings strategic advantage. Most other robot labs, by contrast, favor software over hardware: their emphasis is on programming and simulation, not building actual robots.
“They design things that look great on paper but would never work in real life,” D’Angio said.
Lee, the grad student from Hawaii, stumbled across a video of one of HongÂ’s robots on YouTube and sought him out. As an undergraduate, Lee designed an underwater robot that resembled a five-foot tuna, a means to explore how the tuna swims so swiftly.
Hong started out at a disadvantage when he joined Virginia Tech in 2003. He was a mechanical engineer -- a hardware man -- in a discipline dominated by computer programmers. Many in the field believed the mechanical design of robots was a dead end.
Hong submitted idea after idea, culled from his notebook of diagrams and scribblings, but each of his research proposals in 2003 and 2004 was rejected. He sat in his home in Blacksburg, 7,000 miles from Seoul, and contemplated his future.
The breakthrough came with a machine aptly named IMPASS, or Intelligent Mobility Platform with Active Spoke System. Funded by the National Science Foundation in 2005, IMPASS looked like something built of Tinker Toys.
Hong’s innovation was to remove the rims from a pair of wheels, leaving only the spokes, which could be “taught” to extend and retract in reaction to the surface. The result was a robot that could crawl over rough terrain.
HongÂ’s robot friends quickly multiplied: STriDER, a three-legged robot that resembles an animated camera tripod; HyDRAS, or Hyper-redundant Discrete Robotic Articulated Serpentine, a robot that resembles a Slinky and can roll up telephone poles; and MARS, Multi Appendage Robotic System, a crab-like robot patterned on the NASA rover.
This year, most of the labÂ’s energy has gone to a new project: SAFFiR, or Shipboard Autonomous Fire Fighting Robot. SAFFiR will resemble a more athletic CHARLI, built to help the Navy fight fires on ships. Eventually, he could be worth millions to RoMeLa.
Each of those projects has brought RoMeLa fame. But HongÂ’s greatest ambition, his holy grail, was always RoboCup.
The lab built its first humanoid soccer player in 2004 and called it DARwIn, or Dynamic Anthropomorphic Robot with Intelligence. It looked like a pint-size Terminator. An early video shows DARwIn struggling to its feet, teetering and falling; two students swoop in to grab it before it tumbles off the table.
In 2007, DARwIn was Virginia TechÂ’s first entrant in RoboCup. Hong and his students traveled to Atlanta. Their three-robot team played in the smallest of three size categories. During one match, DARwIn fell, and its head broke off. It stood, headless, and began walking in circles. The team returned home in humiliation.
Hong’s students developed DARwIn-2, with stronger joints and better “eyes.” Some of the software came from an autonomous vehicle designed for Virginia Tech’s entry in the DARPA Urban Challenge, a 2007 competition to spur development of autonomous vehicles sponsored by the deep-pockets Defense Advanced Research Projects Agency. Virginia Tech’s entry, a modified Ford Escape, had finished a surprise third, ahead of MIT, Penn and Cornell. Later, Hong parlayed the DARPA success into the Blind Driver Challenge: His team built the first vehicle designed for blind drivers. The interface worked by sending signals to the sightless driver to speed up or slow down, turn right or left, according to feedback from external sensors.
Hong and the students traveled to Suzhou, China, for RoboCup 2008. There, Team DARwIn finally played a passable game of soccer, inspiring chants of “DAR-WIN, DAR-WIN” from the crowd. But the machines still couldn’t “see” well. When light conditions on the field changed, DARwIn lost track of the ball.
For RoboCup 2009, in Graz, Austria, Hong made the risky decision to install a cutting-edge camera system in DARwIn-3. The camera was no larger than a credit card and insanely complex but amazingly powerful -- when it worked. But the camera failed, and DARwIn-3 froze on the field.
By 2010, years of development finally paid off. DARwIn-4 boasted radical improvements in strength, weight and agility. In Singapore, HongÂ’s robot team finally bested the best robots in the world, amassing four victories.
But RoboCup is a large event with robots of many shapes and sizes. To win the coveted trophy, it was important for HongÂ’s team to show its technological depth by fielding an adult-size robot alongside the diminutive DARwIn. In 2010, Hong entered CHARLI.
There are good reasons why America has struggled to build a true humanoid robot -- an autonomous, human-size machine with hands and feet and a head -- before CHARLI. The simple human task of walking is torturously difficult for a robot. Even today, no robot can mimic the rhythm of human movement, a sort of perpetual falling. Even ASIMO, the fastest humanoid robot around, walks with a distinctly robotic gait.
The first CHARLI, built with $20,000 in seed money by Hong and his graduate students in 2008, walked in a wobbly, drunken shuffle. He was fragile, and he suffered fatal crashes in Singapore. By the end of the tournament, Virginia Tech had four victories and two defeats, good enough for fourth place among 24 teams.
For RoboCup 2011, in July in Istanbul, Hong and his students built CHARLI-2, stronger, lighter and more nimble than his ancestor. In Istanbul, HongÂ’s DARwIn robots easily won their games. CHARLI didnÂ’t fare so well. In his first match, CHARLI kept losing sight of the ball and failed to score a single goal.
Between games, on the sidelines, Hong and the students racked their brains for a solution. They finally isolated the problem: CHARLI had a blind spot over his shoulder. They fixed it by hastily reprogramming CHARLI to approach the ball in a wide arc, so he could keep his sensor “eye” on the ball.
CHARLI charged back in a blaze of goals, setting a tournament record and positing himself as the Pele of robots. It came to a final match against Robo-Erectus, a black, bulb-headed robot from Singapore. The two machines teetered around the field for what seemed an eternity. Finally, CHARLI reared back and kicked a decisive shot past Robo-Erectus, who greeted the challenge by freezing into a submissive bow.
Hong, on the sideline, pumped his fist in triumph. Students leapt for joy. The trophy was coming home.
Back at Tech this fall, Hong was leading a tour for visitors from an entrepreneurial robotics lab in Fredericksburg, Va.
The group rounded the corner to a hallway filled with plaques, framed awards and press clippings. Hong drew their attention to the concrete-block trophy wall. In coming weeks, the RoboCup crystal globe trophy would take its place of honor in a display case upstairs on the engineering buildingÂ’s main floor.
“If you want the best students,” he told the group, “this is the place.”
Gesturing toward the trophy wall, Hong noted to the visitors that it was now out of space.
Source:
No comments:
Post a Comment