Skip to main content
book excerpt: part 2

Based in Bedford, Massachusetts, a short drive north from Boston through leafy New England, iRobot's headquarters is a single building in an industrial park overlooking the freeway. It's a much smaller base than what you would expect for what is quickly becoming one of the world's most important companies. Still, the modest operation speaks volumes about the company and the industry in general-while it is full of promise and has had some success so far, it is still very much a nascent business. iRobot was founded in 1990 by Colin Angle, Rodney Brooks and Helen Greiner, a trio of robotics researchers from MIT, and took its name from the Isaac Asimov novel I, Robot-since turned into a movie starring Will Smith-wherein humans and machines live together in relative harmony (at least before the machines rise up and rebel). In its early days, the company built some impressive robots, including a toy dinosaur, but floundered about in search of a market for its technology. "Probably about fourteen or eighteen business models came along and were discarded as they were found to be little more than a subsistence existence," explains Angle, the chief executive officer. Like many high-tech executives, Angle eschews a suit in favour of casual attire, a simple button-up polo shirt to complement a slightly dishevelled look that he must have borrowed from Bill Gates.



Read part-1 of the 5-part excerpt from Sex, Bombs and Burgers

Opportunity finally came knocking in 1997 when the company designed Fetch, a robot that cleaned up cluster bomb shards from airfields, for the air force. That led to a DARPA contract the following year for the PackBot, a robot that resembles a lawn mower, but with tracks instead of wheels and a long multi-jointed arm sticking out of it. The "platform," as iRobot calls it, can be endlessly customized with whatever equipment is desired. The arm can be equipped with several different types of cameras, including night-vision, as well as additional claw-like "hands" or even sensors for detecting explosives and biological weapons. Weighing twenty kilos and costing $150,000, the PackBot turned out to be a godsend for troops in Afghanistan and Iraq. Its small size and ruggedness allowed it to go just about anywhere, from caves to office buildings, and its customization options let it perform many different tasks, including reconnaissance and bomb disposal, which has come in handy battling the weapon of choice of Iraqi insurgents, the improvised explosive device (IED).

The military contracts gave iRobot financial stability and allowed Angle and his cohorts to think about their real goal: the consumer market. In 2002 the company rejigged the Fetch into the Roomba, a disc-shaped vacuum cleaner that looks like a big Frisbee. At $200, the Roomba was the first home robot that was affordable, smart and-best of all-useful. With the press of a button, the device vacuums a room and returns to its charging station when finished. It can detect and avoid walls, coffee table legs and stairs, and go where humans can't, like under the couch.



The Roomba became a hot Christmas gift and proved a hit for the company, which then expanded into other home robots such as the Scooba floor washer, the Looj gutter vacuum and the Verro swimming pool cleaner. On the backs of its dual military and consumer markets, iRobot turned a small profit in 2003 and has continued growing since. In 2005 the company went public on the NASDAQ stock market and in 2007, before the global recession soured virtually every industry, reported a profit of $8 million. The next year the company sold more than a million home robots worldwide, bringing its total Roomba sales to three million, and saw overall PackBot deployments reach 2,200.7 The two markets have forced the company to learn different lessons, but in the end they are more similar than you'd think. "The consumer marketplace is very, very price sensitive. Everything needs to be engineered in an integrated fashion. There's no opportunity to put any fat into the design and still make money," Angle says. But both markets are utility-driven businesses. "If the Roomba doesn't actually clean your floor, we don't sell them. In the military, if the robot doesn't provide a tangible benefit to the soldier, such that the soldier is demanding to take the robot, then sales don't happen either. It's purely a utility sale as opposed to an entertainment or gadget sale because you don't buy vacuum cleaners on a whim, nor do you lug around a forty-pound robot for giggles."

One of the big differences in the company's two main products is their respective degree of autonomy. The Roomba operates with a high level of independence because the biggest problem that can arise is a tussle with the family dog. The PackBot and the company's other military robots, however, are remote controlled because troops cannot tolerate any unpredictability in their equipment like, say, having it turn on them Terminator-style. This is why autonomy for more sophisticated robots is coming in "on cat's feet," as Joe Dyer, the company's president of government and industrial robots and a retired navy vice-admiral, puts it. Dyer, a former pilot who displays the control stick from his old F-18 on the conference table in his office, explains that even today's super high-tech planes saw all of their automation creep in one small piece at a time, starting with directional stability, then cruise control, then automatic landing. Autonomy for robots will happen the same way. "The next step of this is to say, 'Look robot, how about if you lose communications, you be bright enough to go back to where you can talk to us.' Autonomy really does come in on those cat's feet. It doesn't go to the governor of California."

The pace of that autonomy is likely to quicken over the next few years, since the current system of always keeping a man "in the loop" is tremendously inefficient. On a cost level, the benefits of replacing a human with a robot are voided if you still need that human to control the robot. On the consumer front, iRobot probably wouldn't have sold many Roombas if a human were still required to monitor the vacuum. Militarily, robots still have to transmit everything they see back to their operators, a considerable waste of wireless bandwidth, which is often at a premium in combat situations.

The solution to both issues is to give war robots more freedom by making them smarter. One possible scenario involves a sort of "swarm intelligence" where robots lead themselves. Imagine, for a minute, that a fleet of armed aerial drones are reconnoitering a hostile area when one of them is fired on by insurgents. Sensing that their cohort is under attack, the other drones head to its location to help out with the fight. Once the conflict is over, each individual drone resumes its original position, ready for the next encounter. Such independent swarm-minded robots may sound like the Borg from Star Trek, but they could act much more quickly and efficiently than they would if they were individually controlled by humans, who have to approve every move. Such systems, of course, will come in "on cat's feet" because of the inherent risks involved. "You've got to make sure you get it right because for a whole bunch of reasons, you may not see it for another ten years if you screw it up," says Kevin Fahey, the executive responsible for purchasing robots for the U.S. army.



Fighting the Future

It's developing this sort of autonomy, and doing so at a low cost, that was the whole point of DARPA's road races. In 2003 the agency announced the Grand Challenge, an ambitious 480­kilometre desert road race for unmanned cars between Los Angeles and Las Vegas. The contest was framed as a mini-Manhattan Project, with everyone from advertisers, corporate sponsors, science-fiction writers and even movie producers called on to get involved. "In order to make the DARPA Grand Challenge a success, we must maximize participation by everyone from major players to amateur enthusiasts," the program's manager said in announcing the race.

The prize money was set purposely low at $1 million to encourage entrants to concentrate on cost efficiency-after all, no one was going to spend $10 million to win $1 million- while a desert course was chosen not only in case the machines turned on their human masters, but also because of the terrain's similarity to Iraq and Afghanistan. The inaugural race, held in March 2004, could have been considered a flop, since none of the fifteen participants actually finished-Carnegie Mellon's "Sandstorm" went the furthest, a whopping 12 of the 480 kilo-metres. But the entire idea was redeemed a year later. With the prize money raised to $2 million and the distance dropped to 212 kilometres, five of the twenty-three competitors finished the second race, led by Stanford University's "Stanley," which clocked top speeds of sixty kilometres an hour.

Six of eleven teams finished the third race, in 2007, led by the "Boss," despite the heightened complexity of driving in an urban environment, a city setting recreated on a closed military base in California. DARPA director Tony Tether was ecstatic about how much progress had been made in just three years. "The 2004 event was equivalent to the Wright brothers' flight at Kitty Hawk, where their airplane didn't fly very far but showed that flight was possible ... The significant progress after 2004 was due to the fact that the community now believed that it could be done." Sebastian Thrun, the Stanford artificial intelligence guru who led his team to victory in 2005, summed up the whole affair best: "We all won. The robotics community won."

DARPA's main reason for holding the races was to help it meet a military transformation strategy mandated by the Pentagon. The process started in 1999, after the army experienced yet another embarrassing example of its own bloat and inflexibility. Confused logistics prevented the army from getting its Apache helicopters into Albania for use in the Kosovo war, which led to the conclusion that U.S. forces needed to get lighter and faster-and quickly. The Future Combat Systems plan, which featured a heavy reliance on small unmanned robotic vehicles, was drafted and approved by Congress in 2003. With an estimated budget of more than $200 billion and an expected completion date of 2030, military officials called it the most ambitious modernization of the army since the Second World War and the most expensive weapons program ever. The plan called for the progressive introduction of unmanned air and ground vehicles, starting with a host of reconnaissance and explosives-disposal robots, then eventually moving to machines with weapons. With troops dying in Iraq and Afghanistan, it was easy to rationalize the inclusion of unmanned systems. "If you're at war, the Department of Defense asks for money, Congress moves it around. For the most part, people fund it. They want to support those in harm's way," Fahey says.

The rollout started slowly but ramped up fast. In 2004 American forces had 162 robots in Iraq and Afghanistan; by the end of 2008, they had more than six thousand. At first, the robots came in cave-exploration and bomb-disposal forms (like iRobot's PackBot) and as aerial reconnaissance drones (like Northrop Grumman's Global Hawk). These were followed by the frighteningly named Predator and Reaper armed aircraft, built by General Atomics and operated by remote control thousands of kilometres away on a military base in Nevada. Joining the battle soon will be armed ground robots such as Foster-Miller's MAARS and iRobot's Warrior. While the idea of having robots driving around with live guns scares some people, military experts say the evolution is a natural one that has happened with everything from cars to planes. "Almost every technology that finds itself in military service starts with reconnaissance and evolves to strike," says iRobot's Dyer. "It becomes so frustrating to be able to see but not to act, that it invariably moves to strike capability."

Future Combat Systems wasn't without its critics, however, including President Barack Obama, who inherited the program when he took office in 2009. Obama wasted no time in cutting back on FCS, and while his move was detrimental to big defence contractors like Boeing and Lockheed Martin, it is likely to pay off even more for small robot makers over the long run. Robert Gates, Obama's secretary of defence, announced in April 2009 that the government was cutting spending on "Cold War thinking"-areas of conventional warfare where the United States has a clear advantage-and would instead concentrate on new realities, like fighting the sort of urban-based counter-insurgents found in Iraq. "[Obama's]trying to take resources from areas where we have clear dominance-we control the skies and the seas-and move them to where we're challenged, which is irregular warfare and asymmetrical attack," Dyer says. "Robots are an important part of being able to meet that irregular warfare challenge. You can already see that with the IED threat. It is going to shift resources to an area that is advantageous to the robotics industry and iRobot in particular."

That could mean more programs like the DARPA road races. About 80 per cent of the funding for artificial intelligence research in the United States already comes from the military, a percentage that could increase with the rethinking of defence spending. One thing is for sure: the military's appetite for robots has only just been whetted and is very quickly becoming ravenous. As iRobot's Angle points out, the military market for robots is driving two relatively new concepts into the robotics industry: utility and cost effectiveness. While large Japanese car and electronics companies master the technology and have produced some amazing robots-such as Sony's AIBO robot dog or Honda's humanoid ASIMO-they have so far failed to produce robots that are cheap and useful. "The Japanese industry is a lot about 'cool.' AIBO and ASIMO and dozens of dynamically walking robot projects were developed so that large consumer electronics firms could show they can put together an impressive robot," Angle says. "Why is Matsushita cooler than Sony? Well, just look at their robots. It's bragging rights and it became a huge thing. The population got into it. It's misdirecting the Japanese into a world of show rather than a world of utility." Japanese officials don't disagree. "The U.S. is much better in the commercial field, but in technology, Japan is number one," says Takayuki Toriyama, executive director of the city of Osaka's office in Chicago. "The robotics market in Japan is not expanding right now because they're too expensive. Nobody can afford to buy a robot. This is a very serious problem."

The United States is far from the only big military customer seeking cheap, useful robots. As of 2008, Unmanned Vehicle Systems International, an industry trade group, had 1,400 corporate members in 50 nations while a survey of government-related research found that 42 countries, including Britain, Russia, China, Pakistan and Iran, were working on military robotics. The market for war robots is just starting to open, which means we are still at the early stages of seeing the commercial spinoffs-the Roomba was just the beginning.

Tomorrow: A Smart Slut

While the overwhelming majority of artificial intelligence research is conducted on behalf of the military, some of it is coming from a surprising source: sex-minded criminals. Indeed, if military researchers don't come up with a human-seeming robot intelligence soon, hackers may very well beat them to it.

From Sex, Bombs and Burgers by Peter Nowak. Copyright Peter Nowak, 2010. Reprinted with permission of Penguin Group (Canada)

Follow related authors and topics

Authors and topics you follow will be added to your personal news feed in Following.

Interact with The Globe