|
|
Groundhog
Soccer-playing robots, nursebots, robots that hunt for treasure: A future in which robots increasingly help—or even replace—humans is being realized right here in Pittsburgh, at the Robotic Institute at Carnegie Mellon University. For six years, Lee Gutkind immersed himself in this frenzied world of researchers, graduate students, engineers, and software designers, who must work together—while trying to understand one another—in order to win the global race to create robot autonomy. This chapter, "Groundhog," describes the flurry following the 2002 Somerset mine incident, in which CMU students, led by one wildly passionate, yet difficult professor, convene to create a robot that will be the first to navigate the underworld, with hopes of preventing future human mining disasters.
Somerset is not far from Red Whittaker’s farm, and perhaps sixty miles from the Carnegie Mellon campus in Pittsburgh. Thinking about the near tragedy soon afterward, Whittaker had a sudden inspiration. He began making phone calls to friends in the mining industry, asking questions. Whittaker learned that Quecreek had been flooded because the miners, guided by old, inaccurate maps, had excavated too close to the adjacent long-flooded Saxman Mine, causing the walls between the two mines to collapse. This is a serious problem in the mining industry which no one had addressed-—until now. This is also, quite typically, the way in which Whittaker dives into a project: No money, no plans, no staff, no permission from federal or state agencies to go into the mines he is intending to map. Consumed by his vision and the fire-in-the-belly to think he can make something out of nothing, Whittaker turned to the only means of support remaining--what he terms as his “children’s crusade” or “rookie” option. He immediately designed a course for students and made mine mapping with a robot the course project. The students he attracted, an even mix of engineering and computer science grad students, were talented and, after listening to Whittaker’s idea, motivated. After all, that’s why they came to Carnegie Mellon. This was the school’s reputation: To enable students to get involved in the heat of the robotics action, in the field and the trenches, and to work with and perhaps be mentored by the famed father of field robotics. These young people are the bone and sinew of the institution and the reason the robots and the Robotics Institute exist. Zachary Omohundro immediately bought into the concept of what Whittaker called “technological swashbuckling in coalmines.” Of Scotch-Irish descent, he is tall, slender, and dark, with eyes that seem constantly in motion. “I applied here and at MIT and came to visit both campuses. At MIT, they gave me a PowerPoint presentation about theoretical frameworks for future systems—-and that pretty much made my decision. At Carnegie Mellon there were robots everywhere.” “Robots everywhere” is an accurate description. The High Bay on the ground floor of Newell-Simon Hall, headquarters of the Robotics Institute, resembles a small aircraft hangar. That's where most of the large robots are actually put together and tested--beasts like Hyperion and Nomad, all sitting like warriors resting their thunder, patiently waiting to get back into action. At any given moment, there will be robots here from many research groups, including the Field Robotics Center, the Vision and Mobile Robots Lab, The Robot Manipulation Lab, the Helicopter Lab, the Advanced Mechatronics Lab: Zooming robot Segways learning to play soccer; robots on treasure hunts; Nursebots who care for patients in hospitals; robots who look like Lara Croft (of Tomb Raider fame); robots who attend meetings and give speeches. Robots named Pearl, Houdini, Enviroblimp, Bullwinkle, Grace, Gyrover, Ferret, Demeter, Rhex, Xavier; robots that build other robots, robots that do origami; robots destined for museums as tour guides; robots that squirm along the ground like snakes; robots made from golf carts; robots that creep like caterpillars; robots that are pogo sticks; robots that roll, jump, crawl; robots that resemble beachballs; robots that look like eyeballs; robots that fall apart when they move; miniature robots called milibots—-all zipping back and forth with little regard for pedestrians. Collisions are known to happen. Civilians venture into the High Bay at their own risk. Nearly all of these zipping and zooming little creatures are being controlled by disheveled young men sitting on the floor, on folding chairs, tool cases, portable coolers, huddled over laptop computers, peering at displays with a squinting, manic intensity. Take Chris Baker, a twenty-four-year-old native Pittsburgher, who did his undergraduate work at Carnegie Mellon. Baker is now working on a master’s degree and will soon enter the PhD program in robotics. His father is a general contractor for whom he had worked every summer, so he had certain basic skills--a knowledge of wiring and welding—that the average graduate student lacked. He and Omohundro, who grew up in Minnesota and gained experience building a robot during his senior year at Rice University, were what, according to Baker, Whittaker called “fresh blood": Young people willing to work endlessly under difficult circumstances, which was the gauntlet Red Whittaker put his students through. Baker, Omohondro, and around a half-dozen other students knew that mine mapping with a robot might be the opportunity of a lifetime, although the reality of the experience was, at the very least, sobering. It began with a marathon sleepless weekend of work, starting with a two-hour drive out of the city to a golf cart graveyard. “It was just like a car junkyard,” said Baker. “There were golf carts in many states of disrepair. We selected a golf cart, threw it in the back of a truck, took it to Red’s farm, and tore it apart.” They worked nonstop, alongside Whittaker, over the weekend. “We declared success,” said Baker, “when we hooked it up to the batteries and the golf-cart-now-robot moved.” They named their new mine-mapping robot Groundhog. “Then we could go home and go to bed.” This was only the beginning of the task of turning a golf cart into a working groundhog; weeks of designing and redesigning, writing and debugging software, tearing Groundhog back down to its bare frame to strengthen and rebuild it-—thousands of tedious, sleepless hours. Red Whittaker, like George Patton in World War II, ignited among his troop of rookies a nearly maniacal dedication, as well as an overpowering exhaustion. I have had many conversations with Whittaker. When we talk one-on-one, he sometimes rambles, stumbling over his words and going off on tangents that seem pointless. But groups of young people magically transform Whittaker. A dynamic articulation suddenly emerges. At a class I attended during the time Groundhog was being tested, Whittaker attempted to convince his weary, overworked students to write a proposal for a Groundhog-related project to take place after Groundhog had been tested.
“Nobody around this table looks particularly bored with life and needs something more to do,” Whittaker commented, pausing, sweeping his piercing gaze around the table, peering intensely into each and every eye. (Describing this look of deep scrutiny, one student said it is “like Red is grinding himself into your soul.”) “You don’t have to do it,” Whittaker continued, now staring out the window into the distance, seemingly visualizing the future. But then he leaned forward, wrinkled his brow and raised his forefinger in warning: “Oh, by the way. The opportunity won’t come again. This is your moment.” He paused dramatically. “What are you going to do with it?” This is the Whittaker style and approach, building up a project and then challenging his mostly male students (all male in this case), counting on their testosterone and their competitive, overachieving personalities to take the bait. After discussing the project for a few more minutes, Whittaker toned down his rhetoric, talking quietly, almost soothingly. Then he lapsed into a long silence and stared dreamily into space. “Red’s way,” said Chris Baker, “is to say, ‘We need somebody to do something,’ and then there will be this really intense quiet until somebody says something to break the silence.” Whittaker was usually willing to outwait the discomfort he had caused. Whittaker, according to Omohundro, has an “optimistic zone” around him. “Once you get into it, it is kind of hard to disagree with him.” He is “inspiring,” said Omohundro, “and terrifying.” Baker and Omohundro, along with two or three other students, built Groundhog from scratch. “Red has an obsession that both Chris and I understand,” said Baker. “People who understand tend to gravitate toward him while people who don’t tend to drift very far away.” Indeed, many of the graduate students and PhD candidates I talked with have come to Carnegie Mellon at least in part because of Whittaker’s legendary status. But once on campus, they can be put off by his unrelenting expectations. His aura is magnetic, but his presence is frequently avoided, either because his moods are unpredictable—-he can lash out unexpectedly-—or, worse, he can charm or manipulate you without your ability to resist. “Meetings with Red don’t entail conversation,” a Carnegie Mellon graduate engineer who has worked with Whittaker told me. “They are more like motivation lectures. Anytime I talk with him, it feels like he is trying to figure out how to get the best results from me. So if criticizing--hammering on you--gets the results, or telling you that you are great is going to get the best results, then that is what he will do."
Neither Baker nor Omohundro were unaware of the significance of their hard work and the purpose behind it: “Sometimes we forget the fact that no one has ever put a robot in a mine before. We are students and we are making history,” said Baker. “That’s awesome!” Making a robot operate autonomously in a mine presents special challenges, perhaps even more perplexing than operating one in outer space. Robots are more effective on level ground, in an open environment with good uniform lighting. GPS (global positioning systems) or satellite imagery will help position and guide a robot-—even in the Atacama or on Mars. But an unmapped mine is a dark, unresponsive mystery. “The robot sensors are less reliable underground,” Aaron Morris explained. Morris is short, solidly built, and strikingly handsome with a pale, baby-faced complexion. His words are laced with the flat, guitar-like West Virginia twang where he grew up, although, ironically, he had not ever entered a mine before the “subterranean” robotics course. But he had learned a lot about his home state since entering Carnegie Mellon and the integral relationship robots might have with the mining industry. Since it was an exploration task--the mine into which the robot entered is a black box-—there was no way to plot where the robot was going or anticipate what it would encounter: fallen timbers, flooded shafts, abandoned machinery. Morris explained, “You program a robot to think through a problem and make intelligent, accurate decisions. Mistakes are often irreversible. A robot lost in a mine can usually not be rescued and may never return.” Groundhog sees by building a two-dimensional map of where it is going, called a “cost map,” meaning that, as it travels, the code allows it to see the obstacles in its path and attempt to avoid them or analyze the potential cost of going forward and over or under them. Robots will not recognize an object it sees. A harmless cable hanging from the ceiling and a dangerous pillar may look roughly the same to Groundhog. The word “see” doesn’t accurately capture the reality of the robotic vision and navigation system, by the way. Groundhog will find its way into and, if all goes well, out of the mine by counting and memorizing intersections. “That’s what you, anybody, would do if you were in a cave. You come to an intersection, and you are immediately at a decision point-—left or right or straight. If you go straight, you will leave a marker and therefore when you return you know you have been there before.” I asked Morris about the maps I had seen Groundhog and other robots make—-vivid, colorful illustrations of the inside of mines or other areas it or other robots may have traversed. “We build a 3-D model after the fact for easier viewing, using the 2-D data the robot has gathered.” The maps, he stressed, are for people to view--not what a robot will see. Veteran roboticists at Carnegie Mellon, like Whittaker and his protégé Wettergreen, insist that they do not feel any connection to the robots or the code they conceive and create. But rookies like Morris can’t identify with the ambivalence of their teachers. “Red likes to say that humans don’t have a personal connection to the robot—-but he is wrong-—there is always a personal connection to the robot. Code is a reflection of the person who writes the program, so how could I not feel connected? These robots are extensions of us. If they don’t do what we program them to do, we are going to feel a great sense of failure.”
Morris smiled and shook his head slowly back and forth, weighing his response. “This is something I ponder all of the time, and I don’t really have an answer. ‘Thinking,’ ‘intelligence,’ cognition’ are such vague words. Programmers writing code lay down a set of rules for a robot to follow, so maybe the robots are simply following orders.” But then he paused to reflect and shrugged. “But who’s to say that thinking, for all of us, is not following a set of rules? People are consistently determining and breaking boundaries.” But “thinking” or following rules like people may not always be the objective, many roboticists contend. Robots have sensors, global positioning systems, gyroscopes; they often know a lot more than people about a particular situation or problem and can jump start the cognitive process. To maximize efficiency, humans and robots may very well need to think differently. Morris did his undergraduate work at West Virginia Institute of Technology, an independent school recently absorbed by West Virginia University. He started out in engineering because “[his] father thought it was a good idea.” But he moved to robotics as a graduate student at Carnegie Mellon, and now after five years of study, he is working on his PhD. He does not consider himself a typical robot-rookie, however. “After five years, I have become an old man here.” He is twenty-six years old.
Whittaker smiled. “That’s one voice. Anybody else?” Four more hands were raised. “Five of nine, that’s terrific,” he proclaimed. Then Whittaker hunched over and peered intensely into space, foreseeing the future and sealing the deal by anointing his players. “I actually pity anyone who is competing against you because . . . who could run with you? Who could touch you?” He went on like this for a while, then asked: “How do you get thirty good pages?” The proposal must be thirty pages long. "Lots of pictures,” one student yelled out. Whittaker laughed: “How do you get thirty great pages?” “Write the first one,” another student replied, quoting his mentor. “Ok, is there anything else to say?” Whittaker waited. No one answered. “Going once, going twice . . . Who is leading?” “I guess I am,” a staff member mumbled. “If you have a sense of conviction, you don’t have to guess,” Whittaker replied. Whittaker waited again, but the specter of yet another exhausting, never-ending commitment was now, suddenly, upon the “new generation,” and they were, perhaps, momentarily paralyzed. “Ok, make it work,” Whittaker declared, as he jumped from his chair and dashed out of the room. Later, when I asked Aaron Morris how he felt about being confronted with yet another daunting job to do, he shook his head, spread his arms as if feigning helplessness, and laughed: “We raise our hands. We follow. We take that gauntlet. We like to do what people say is not possible.” As it turned out, the proposal they committed to write on that day was written and made the deadline--and then was rejected. But according to Aaron Morris, “The hard work, even though it was completely over the top, kept us focused.”
from Almost Human: Making Robots Think by Lee Gutkind. Copyright © 2006 by Lee Gutkind. Used by permission of W.W. Norton & Company, Inc. TABLE OF CONTENTS | ABOUT THE PROJECT | EVENTS | CNF WEBSITE |
In July 2002, the world was transfixed on the tiny town of Somerset, Pennsylvania, as rescuers drilled into nearly 300 feet of dirt and rock in a desperate attempt to save nine miners trapped underground in the flooded Quecreek mine. For a while, rescuers didn't know if the miners were alive until they heard tapping from survivors—-signals coming from what turned out to be a small air pocket. After seventy-eight hours of ceaseless digging and pumping, all nine men were rescued. Their stories have been told repeatedly and captured in a book and a made-for-TV movie. Their plight and bravery affected anyone who followed the rescue, most especially Red Whittaker.
The students, clustered around a long rectangular conference table, listened quietly, hardly moving, as Whittaker described the project and all of its potential benefits. It was difficult to know if they were enthralled by his rhetoric or stunned by his presumption. But they were clearly in his grasp.
But Whittaker gets a tremendous amount out of the people who choose to remain with him, despite his tantrums and manipulations. Omohundro, Baker, and many others who are swept up by a Red Whittaker project talk in terms of “thirty- and forty-hour days.” They mark significant moments in the project by saying, “That’s when we could go home and sleep.”
"Are you thinking for robots when you program—-write code-—for them, or are you helping them think?” I asked. This is a key aspect of robotics, and a subject of constant debate and reflection. Are the robots simply reflecting the programmer’s wishes—-or, when they traverse a mine (or Mars) are they acting on their own?
But, like his fellow students in this conference room listening to Red Whittaker attempt to motivate them, he couldn’t seem to focus on much of anything else. “We eat, sleep, and breathe robotics,” he said. “I wander into a convenience store and think of all of the ways robotics will be important. We become obsessed,” he said, waving his arm like a flag. “We aspire to change the world.” This seemed to be a quality—-or a malady—-that most of the rookies not only under Whittaker’s spell but also throughout the Robotics Institute shared. Which is why Morris was the first to raise his hand in the classroom that afternoon when confronted by Whittaker’s new proposal. “We have to do it,” Morris said.
