E-Mail This Article to a Friend Deus ex machina (continued)

THE WORD "robot" was coined in 1920 by playwright Karel Capek. Its derives from the Czech word robota: "drudgery." And indeed, for decades robots have been used to perform the sorts of repetitive, unpleasant, sometimes dangerous tasks humans would rather avoid. But from the beginning, humans expected robots to do and be more. Science-fiction books — like Isaac Asimov’s classic I, Robot, from which iRobot takes its name — are rife with walking, talking automatons who look and function like people. But aside from a few laboriously constructed specimens — Asimo, for instance — real-life robots like that are anomalous. It’s simply too difficult to implant the nuances of mobility, perception, and cognition into a functioning machine. Even something as seemingly simple as walking with fluid motion poses a problem scientists have spent decades trying to solve. The explosive development of computer and information technology in the past decade-plus has only stoked the appetites for robots who are more like us. But despite sure steps forward, robot evolution has, by and large, been slow.

"The first thing out of everyone’s mouth is, ‘When are you going to build me Rosie [from The Jetsons]?’ " says Nancy Dussault, director of product marketing for iRobot’s Consumer side. "When am I going to get a robot in my house that’ll do everything for me?’ There’s been a lot of progress, but it’s going to be a long time before you see that type of technology in day-to-day life."

Lately, however, there’s been a shift in thinking, one sparked in large part by people like those at iRobot. "There’s been a fundamental change in philosophy," Angle says. "Maybe 10 years ago, the notion was a top-down approach, trying to understand how human cognition works." (The great successes in this area were chess-playing programs, like IBM’s Deep Blue.) "And, largely because those problems are just too hard, all of that work failed. Where success is happening today in the industry is that people are being more opportunistic. They’re saying, ‘Hey, I have a great algorithm, which, if you give me a face, will differentiate that face from another face.’ It’s not going to tell me if that person’s happy or sad, or even extract anything recognizable that you think about when you think about assigning a face to a person. But it can tell faces apart. That’s opportunistic and that works. There are other visual systems that can track various elements, that can distinguish hallways from rooms, start to differentiate one room from another room. This is done in ways that are difficult to explain. But it works."

iRobot scored its first government contracts, primarily from the Pentagon’s Defense Advanced Research Projects Agency (DARPA), in the early 1990s. But it wasn’t until the 2002 Afghanistan campaign that the company’s robots were deployed to any real battlefield situations. The PackBot, years in development here at home, immediately proved itself up to the task. "[DARPA] weren’t sure what they wanted on the robot, so we came up with this idea of having all these payloads, ports, areas where you can add sensors and different things to it. It can expand as they come up with different needs," says Tom Ryden, who oversees sales and marketing for iRobot’s G&I division. "Our focus is really making a rugged robot. Before this, robots were more ‘lab queen,’ more used by researchers as experiments. They couldn’t take the daily abuse soldiers would subject it to."

Seeing the PackBot in action, it’s striking that something so expensive — units cost between $50,000 and $100,000 — and complex is meant to be thrown through windows and tossed into the backs of pickup trucks. (It can withstand a two-story drop onto cement.) Ryden stands over the PackBot’s control center, a surprisingly simple interface — two video screens, arcade-style disk-shaped controllers, and assorted other levers and knobs — housed in a sturdy metal suitcase, and explains how it all works. "It’s a lot like a video game," he says. The screen shows two camera shots. One, meant for steering, is a fish-eye view of the ground in front of it. The second, which is much higher-resolution, comes from a camera on top of the PackBot’s arm. There’s also a screen showing a three-dimensional computerized rendering of the PackBot itself, which allows the soldier manning the controls to see what position it’s in, even when it’s out of visual range.

Gingerly manipulating the claw-gripper at the arm’s elbow, Ryden demonstrates just how delicately it can be controlled as he pilfers a co-worker’s jacket, grabbing it by the coat loop from a hook on the side of his cubical. It takes a couple of tries, but the arm is amazingly agile, sensitive to the slightest touch; it moves much like a human arm, with all the degrees of torsion — something Ryden says took quite a bit of work to sort out. "To actually get the robot to do that and have you control it is hard to do," he notes. "We’ve seen a lot of advances in that area as well."

THE PackBot is iRobot’s flagship military robot. But the company is working on many others, most of them still in the research-and-development phase. "We’ve got some robots that are inspired by nature," Ryden says. "We developed a robot that could operate in a surf zone — where the waves crash on the beach is a very hard environment to operate in. So we developed one that’s actually very crab-like. It has six legs, the legs invert, and it can kind of crawl through that area. Again, the idea was to be able to examine for possible mines. We also did a wall-climbing robot that was inspired by a gecko."

The SwarmBots, which are still in the research stage, are hundreds of tiny, five-inch robots that mimic the communal actions of ants and bees, communicating with each other so humans don’t have to. "Throw 50 of them into the subway if you think there’s a chemical scare," says Dussault. "One of them says, ‘I found something over here that might be a chemical,’ and calls over the robot with the chemical sensor. The robot detects it: ‘Yes, we have anthrax.’ And then the other robot comes over and says, ‘It’s affected this area.’ They sentry the area. And then one robot that hasn’t been affected will go up and communicate to the base that we’ve found anthrax, in this corner, here’s the location, and it’s being contained. All without a human having to go into harm’s way."

This brings up an interesting point. We live in an era where terrorism is an omnipresent threat at home, and where we’re currently engaged in two wars abroad. Surely, bad as they are, times like these must be good for business. "Yes, in the sense that there’s more of a demand for robots," Ryden says. "But when we developed these robots there was never a thought that we were developing them for war purposes. We developed them for helping soldiers in many areas. Still, we were very thankful that when the war came we were able to help and get some of these robots out and get experience. The more experience we get overseas in actual uses, the better we can make these products and make them more advantageous for the soldiers to use in the future."

Another machine iRobot is developing to do just that is a larger, one-ton vehicle, similar in utility to the PackBot. "You’re limited with the smaller robots how much you can pile on," Ryden says. "There are a lot of convoys in Iraq, and convoys are very dangerous. You could have one vehicle be manned and the rest be autonomous and just follow. It relieves some of the normal work for soldiers, makes it safer. Those are the areas we’re interested in. We’ve already seen the [unmanned aerial vehicles] take that step, and when they first came out it was a slow acceptance. But once the military got more confidence in the unmanned air vehicles, you hear about them a lot more. As the military gets more accepting you’ll have more ground-based vehicles: perimeter patrols, robots making sure no one’s getting on the military base. And that will expand to homeland security, border patrol, port patrol. You’ll see a lot of unmanned vehicles being able to do those tasks."