PREPARE yourself for a robot invasion. In the next few years robots will be moving into your home and taking over your life. But don鈥檛 worry: they want nothing more than to please you. The only danger these bots pose is getting under your feet as they busy themselves around the house. Humanity is about to domesticate the robot.
Can it be true? After years of hype and disappointment can we finally put our feet up and leave the housework to tireless mechanical servants?
Well, you can already buy a robot that mows your lawn while you recline in a hammock. Later this year you鈥檒l be able to buy a cleaning bot with enough intelligence to vacuum that cluttered space you call your living room (see Diagram). Sure, it鈥檚 not an invasion quite yet. According to a recent report by the United Nations Economic Commission for Europe (UNECE), a mere 3000 household robots had been sold worldwide by the end of 1999. But wait. The same report predicts that by 2003, there will be as many as 290,000 robots beetling about our homes. And that doesn鈥檛 include the flood of robotic toys, such as Sony鈥檚 Aibo dog and Hasbro鈥檚 My Real Baby doll, that is arriving on the market.
Advertisement
Soon, robots will be taking out your trash and mopping your kitchen floor. They鈥檒l be followed by multi-purpose domestic servants that can fetch you a beer, pick up your dirty laundry, water your plants and keep an eye on your elderly relatives. Eventually robots will have enough intelligence and common sense to obey voice commands or gestures. Never again will you have to wait in for a plumber to fix your toilet. A maintenance robot will do it for you-just as soon as it鈥檚 finished servicing your car.
It鈥檚 an old and cherished dream. Ever since the Czech playwright Karel Capek coined the term 鈥渞obot鈥� in 1921 (鈥渞obota鈥� is Czech for 鈥渄rudgery鈥�), predictions have abounded about how automatons will free us from the daily grind.
In one place they already have. The first domestic robot went on sale in 1966. Called the Aqua Queen, it was a mechanical device that crawled along the bottom of backyard swimming pools, scouring the tiles and filtering the water. Each time it hit a wall it bounced off in a different direction. Its manufacturer, Aqua Vac of Florida, is now one of a dozen or more companies making and selling pool-cleaning robots. The most sophisticated models will spend 20 minutes or so memorising the layout of your pool and then clean it from top to bottom in the most efficient way, including walls and steps.
These pool-cleaners are real robots. They鈥檙e mobile and autonomous, and they can sense if they鈥檝e strayed out of the water. Just pop one in your pool and retire to the sun lounger.
But until the late 1990s, pool cleaners were the only domestic robots in town. Why did it take so long for them to crawl out of the swimming pool and into our living rooms? The reason, to borrow a bit of robotics jargon, is that pools are a 鈥渟tructured鈥� environment. In other words they鈥檙e predictable and uncluttered. It doesn鈥檛 take much intelligence to find your way around what is essentially an empty room with no doors. But transfer a pool-cleaning robot to a less orderly environment and its limitations are cruelly exposed. Swedish company Weda, for example, has modified one of its robots to clean covered reservoirs, but it only works with a guiding human hand.
Robots like these would never cope in a house. In fact, says Rodney Brooks, a roboticist at MIT and co-founder of Boston鈥檚 iRobot company, the home is just about the least structured environment imaginable. Robots are only useful if they can zip from A to B without crashing into things or getting lost. But furniture never stays in exactly the same spot. Doors open and close. Lights go on and off. Things get moved. People and pets roam around.
Up to now, this has always put the kibosh on household robots. In the early 1980s, around the start of the home-computing boom, hundreds of entrepreneurs decided that robotics would be the next big thing and set up companies to make it happen. It was not to be. 鈥淏y 1986 there were only 12 remaining,鈥� says Mark Tilden, a roboticist at Los Alamos National Laboratory in New Mexico.
For a humble 1980s robot, negotiating an unstructured environment was a nightmare. Neither its hardware nor software were up to it. Despite heroic attempts, even the best robots could easily get lost in an empty corridor.
This time round it鈥檚 different. Thanks to a combination of technology and economics, it鈥檚 now possible to build mobile, autonomous robots with enough nous to be useful. It鈥檚 also possible to build them without spending a small fortune. According to Brian Friedman of Robotic Ventures, a venture capital company in Chicago, domestic robots now make commercial sense. 鈥淚t is time for the business sector to get involved in robotics at a new level, in the same way that it helped make the PC and Internet ubiquitous,鈥� he says.
What鈥檚 made all the difference on the technology side has been the exponential growth of computing power, says Shai Abramson, co-founder of Israeli company Friendly Robotics. With the number of transistors that can be crammed onto a silicon chip doubling every 18 months, a modern microprocessor can perform about a billion instructions per second, compared with a paltry million in the late 1980s. Microprocessors are also cheap. 鈥淎 robot in 2000 in nominal terms cost less than half what it would have cost in 1990,鈥� says Jan Karlsson, author of the UNECE survey. But if you take into account the quality of performance, he says, you are only really paying one-fifth of the price.
With all this processing power at their disposal, roboticists have created artificial intelligence systems that can navigate complex environments and adapt to changing situations. Feeding these systems information are arrays of sensors, including sonar and infrared rangefinders-electronic 鈥渃anes鈥� robots use to prod their way around (see 鈥淪how me the way to go home鈥�).
More recent is the advent of tiny motion sensors called micro-electromechanical systems (MEMS). These chip-sized components include gyroscopes, tilt sensors and accelerometers which help the robot tell left from right and up from down. In the early days roboticists would build these from scratch, wiring together transducers and circuitry to make them work. These days you just buy them off the shelf and plug them in.
MEMS are made in a similar way to silicon chips, so they鈥檙e cheap. And because they鈥檙e small they help keep mass and energy consumption down-an obvious advantage in a mobile device. Better batteries make a difference too. Nobody wants a robot that runs out of juice after ten minutes. 鈥淭he prices keep falling and capacity keeps getting larger,鈥� says Carolyn O鈥橞rien, president of Pittsburgh company Mobots.
Thanks to these developments, robots are knocking on the front door. But if you envisage an all-purpose butlerbot complete with bow tie and handy brush, you鈥檒l need to wait a while. The first domestic robots are highly specialised. Friendly Robotics鈥� RL500 Robomow mows lawns, full stop. Husqvarna鈥檚 Auto Mower does the same thing and nothing else. Dyson鈥檚 DC06 does the vacuuming, period. And there are more single-purpose robots to come. Friendly Robotics is working on a leaf collector, a snow blower and a trash carrier. And this year will see the launch of Convenience Companions, a spin-off from Procter & Gamble which will sell specialist floor-cleaning bots. Seems like if you want to hang up your apron for good you鈥檒l need a fleet of robots, one for each chore.
An all-purpose robot won鈥檛 materialise any time soon. Creating a machine to do lots of different chores is a hugely complex problem akin to building a person, says inventor James Dyson. It requires genuine intelligence and integration of a host of skills, most of which no robot has so far mastered. And don鈥檛 forget that cutting-edge technology costs a packet. 鈥淎 $50,000 robot that does everything is not going to sell,鈥� says Brooks. It would be cheaper to hire a cleaner.
As it is, today鈥檚 manufacturers struggle to keep their prices down. Robotic appliances generally need more components and cost more to develop than their manual ancestors. While you can pick up a lawn-mowing robot for around 拢500, a Dyson vacuum robot will set you back 拢2000, ten times as much as the company鈥檚 cheapest manual model.
Yet the manufacturers are confident that prices are low enough to attract the 鈥渆arly adopters鈥� who鈥檒l get the ball rolling. 鈥淚鈥檓 not committing commercial suicide,鈥� says Dyson. 鈥淚 believe that in five years most of us will be using robotic vacuum cleaners.鈥� In time, he says, prices will come down and most consumers will be able to afford one.
To keep costs down, most of the first domestic robots on the market have been built from the cheapest components. The irony is that rather than being state-of-the-art, today鈥檚 lawnmowers and cleaning bots use technology invented in the early 1990s. Not for them the very latest 64-bit microprocessors, instead they run happily on 16 or even 8-bit microprocessors, some of which have been around for eight years. Another trick is to use engineering rather than expensive sensors. Why give a floor-polishing bot a vision system to detect when it has strayed onto the carpet when you can simply measure the drag on its wheels? With shortcuts like this Convenience Companions reckons it can sell cleaning robots for as little as $30.
What鈥檚 more, far from being intelligent devices capable of learning or knowing when their services are needed, the majority of today鈥檚 domestic robots are quite stupid. None use cutting-edge AI. Most even lack navigational skills, preferring instead to mow or vacuum at random within the confines of a room or perimeter wire. Only two can make it back to their charger when their battery is running low.
For research roboticists eager to show off their latest AI technology, working under such constraints can be frustrating. 鈥淲e work to a spreadsheet where there is a bottom line on what we had to meet,鈥� says Brooks, whose iRobot company worked with Hasbro on the interactive doll My Real Baby. The goal was to make the doll as lifelike as possible, but if a component cost a penny too much the bean-counters vetoed it-even if it would have made a big difference to performance. 鈥淭his almost made some of the engineers cry,鈥� Brooks says.
But there is a positive side to this penny pinching. Keeping the robots simple means less can go wrong. That means fewer malfunctions and safer performance (see 鈥淪afe as houses鈥�). It鈥檚 a question of not setting yourself up to be knocked down, says Friedman. 鈥淚f you keep it simple and make it cost-effective people will be willing to use it and accept that it won鈥檛 clean in the same way as you,鈥� he says.
So how do roboticists plan to make their creations more versatile? To begin with, says Dyson, we鈥檒l see simple add-ons with, say, your vacuum cleaner being adapted to detect fires. But eventually we鈥檒l get robots that can cope with a whole variety of chores around the home.
Some companies are already heading in that direction.
Gecko Systems of Georgetown, Texas, for example, has started selling a general-purpose robot called CareBot. Inventor Martin Spencer describes it as a 鈥渂asebot鈥�-a platform onto which he鈥檒l build useful functions.
The clever thing about CareBot is that instead of having its own 鈥渂rain鈥�, it taps the spare processing power of a PC via a wireless link. That means the robot itself is cheap and light but can still access cutting-edge computer chips. At the moment all CareBot can manage is a spot of vacuuming, but Gecko Systems is teaching it other skills, including smoke detection and plant watering. CareBot will soon get a vision system and voice recognition software too, making it even more versatile. Add an Inspector-Gadget-style arm that can pick up clothes, cans and books and you鈥檙e not that far from a butlerbot that can follow complex orders such as 鈥渢ake John a beer鈥�.
Using a PC as the robot鈥檚 brain has other advantages, says Henry Thorne of Probotics, a Pittsburgh-based company. Probotics makes a multi-purpose robot called Cye, which comes with a vacuum attachment, a drinks trolley and a webcam. You give it orders through a software package complete with icons and drop-down menus. That means anyone familiar with Windows-based computing can work it, says Thorne. 鈥淢y ten-year-old child can use it,鈥� he says. 鈥淚f it were text-based he couldn鈥檛 do that.鈥�
Karlsson agrees that ease of use is vital for the home market. Until now, robots were only common in industry, where the competence of the operator could be taken for granted. Consumers are less willing to invest time learning how to master something. They want to take it out of the box and use it straight away.
Probotics鈥� midterm goal is to have Cye fetch a beer from the fridge, says Thorne. This may sound simple, but it鈥檚 quite a complex task. The robot has to find the fridge and open it, locate a beer and pull it out-without dragging egg boxes, bags of salad and mayonnaise jars onto the floor. 鈥淭hat鈥檚 something I know I can build,鈥� Thorne says. 鈥淚n 15 years I want it to be able to fix your toilet.鈥�
At the moment Cye is little more than a toy. But that may be an advantage if it helps consumers accept it more readily. 鈥淭he expectation of toys is much lower,鈥� says Brooks. 鈥淭hey don鈥檛 have to achieve any task other than entertainment.鈥� For that reason, some industry observers think that toys will be the first domestic robots to take off. In the 1970s many people resisted the idea of a home computer, but as games like Pong materialised so did the notion that computers could be fun. If robots follow this pattern then toys like Aibo and My Real Baby could be the trailblazers that make robots a familiar part of daily life.
What鈥檚 more, roboticists are starting to realise that homes are the ideal place for their creations-even more so than industry or the military. 鈥淕etting a robot to be 90 per cent effective is reasonably easy,鈥� says Illah Nourbakhsh of Carnegie Mellon University鈥檚 Robotics Institute in Pittsburgh, Pennsylvania. 鈥淲hereas getting it to 100 per cent is a significant effort.鈥� If a robot asks for help in the home it could be seen as cute, but in industry it鈥檚 intolerable.
As the momentum builds, so too will the race to build increasingly versatile robots. It won鈥檛 be long before domestic robots conquer that most intractable of household obstacles, the staircase (see 鈥淭wo legs better鈥�). But on the intelligence front, perhaps we shouldn鈥檛 expect too much. As the old AI joke goes: if you could build a robot intelligent enough to do the ironing, it would be intelligent enough not to want to do it.