The build-up Good and ready After slow beginnings, a big push in robotics now seems imminent

The build-up

Good and ready

After slow beginnings, a big push in robotics now seems imminent

ON THE OUTSKIRTS of Odense, a small city in southern Denmark, Enrico Krog Iversen shows off the building he has bought to serve as the new headquarters and assembly facility for Universal Robots, a company of which he is both the chief executive and a big shareholder. It is about ten times the size of the company’s current headquarters, a five-minute drive away. Universal Robots, founded in 2005 by academics from the nearby university, is growing pretty fast. In the past four years, says Mr Iversen, its sales have increased more than 40-fold. By 2017 he hopes for a turnover of DKr1 billion ($190m).

Universal Robots makes robot arms that are light and easily programmed, and hence well suited to use in small manufacturing businesses. At €22,000 ($31,000) each, plus a similar amount in set-up costs, they are also affordable. Universal’s website is stuffed with case studies to demonstrate to potential buyers that the robots’ cost can be recouped in less than a year.

The advent of robots that are cheap and safe enough to be used outside big factories is one reason for a resurgence of interest in robotics over the past few years. Rethink Robotics, a Boston-based company founded by one of the most respected researchers in the field, Rodney Brooks, has attracted a lot of media interest because it sells a particularly appealing and innovative two-armed robot, Baxter, designed for this market.

Whereas Rethink and Universal focus on smallish customers, others are planning to go big. Foxconn, a Taiwanese company that manufactures and assembles electronic kit, says it is aiming to roboticise much of its operation with hundreds of thousands of its own relatively cheap Foxbots.

Affordable does not necessarily mean simple. UBR-1 is a robot arm sitting on an autonomous body that can navigate from place to place. It was developed by Unbounded Robotics, a four-person startup which, like a number of others, was recently spun out of Scott Hassan’s Willow Garage, the hothouse that developed the PR2 robot mentioned in the previous article. UBR-1 is a sort of pared-down, one-armed and less capable PR2, if a much more attractive piece of industrial design. But then, starting at $35,000, it is only about a tenth the price of its older sibling.

This is all a long way from the high-value industrial-robot market served by big manufacturers like Germany’s KUKA, Sweden’s ABB and Japan’s Fanuc and Yaskawa. According to the International Federation of Robotics (IFR), between 2008 and 2012 industrial-robot sales increased by 7% a year, to $8.7 billion. The business is concentrated in Japan, South Korea, Germany, China and America, and on specific industries like cars and electronics. Car companies use the lion’s share of industrial robots; in 2012 they accounted for 52% of robot installations in America. The country with the most robots per person is South Korea, which takes the technology very seriously.

Sensing progress

Cheaper robots should be able to move into fields in which today’s big beasts have shown only passing interest, such as food processing. But that is not the only reason for the buzz around robotics. In academia, robotics, which has been a slow developer compared with booming areas like biotechnology, is on a roll.

The most obvious spur to progress has been the increasing amount of computing power and sensor technology that can be bought for a given price. “Everything built out of silicon is taking off,” says Chris Atkeson, a researcher at Carnegie Mellon. Many of the benefits come in the clever things that people operating in bigger markets have found to do with better, cheaper chips, which can be useful to robot-makers.

Take the Kinect sensor developed at Microsoft for the company’s Xbox 360 game console, released in 2010. An array of microphones and cameras are artfully combined with high-powered chips and well-tailored software to sense players’ movements from a distance and apply them to the game.

Used in robotics, the Kinect sensor is a cheap, easy and fairly reliable way to provide both a sense of depth and a kind of “person-detector”, which is a great help to robots that need to map and navigate their surroundings. Most robot laboratories have some version of it, either bolted to a robot or mounted on the wall or ceiling. A shopping mall in Osaka is wired up with the sensors set up by Japan’s Advanced Telecommunications Research Institute International to tell robots where they are and how to spot the shoppers to whom they are learning to give leaflets and directions.

As well as having access to better sensors and processors from other fields, robotics has devised its own new ways of making software for them. The PR2 was a test bed for the Robot Operating System (ROS), a uniform way of passing messages between the various software routines that run a robot. ROS, now looked after by a not-for-profit spin-off from Willow Garage, the Open Source Robotics Foundation (OSRF), is free to use and easily customised, and is being taken up by more and more researchers, many of whom happily share their fixes to the software. Using an ROS navigation “stack” and a Kinect, it is now relatively easy to build a rudimentary robot capable of finding its way round a building—call it a “trundlebot”—though making it robust and reliable is a lot harder.

S.K. Gupta, a robotics researcher at the University of Maryland who is currently running the National Robotics Initiative at America’s National Science Foundation (NSF), sees ROS and the like not just as solutions to specific problems but as developments that are reshaping the field. Robotics used to be hard to do because to make even a poor robot you had to be good at a whole lot of different things: artificial intelligence, building manipulators, engineering joints and wheels, electronics and so on. As a result, academic robotics research has generally been concentrated at universities that already have a flourishing robotics programme with capabilities across the board, such as Carnegie Mellon, MIT and the University of Tokyo. Now a small team with a fresh insight in a single area—making hands, say, or machine-learning—can use ROS and reasonably cheap hardware to put together a robotic system on which to try out its ideas without being expert in any of the other areas involved. Perhaps as a consequence, the first funding round for the initiative Mr Gupta oversees produced applications for $1 billion in grants, more than 20 times the amount eventually awarded.

Sometimes there is no need to build a physical robot at all. Many of the teams that used Boston Dynamics’ Atlas robot at the DRC had no real-world experience in getting bipedal robots to nip around without falling over. They were chosen through a preliminary competition in which the software they had developed for the trials was used to animate a virtual robot in a simulated environment called Gazebo, which had been developed by the OSRF and hosted in the cloud. Being able to try out robot software in real time this way, says DARPA’s Mr Pratt, is a big deal. He draws an analogy with the early days of integrated circuits on silicon chips in the late 1970s. At first the only way to see how well a circuit was going to work—if at all—was to build it. It was only when simulators became available that designers could ensure in advance that the circuits would work, which vastly sped development in the field.

There was very little academic research in robotics before there were undergraduates who had seen “Star Wars” as kids

Like Universal Robots’ premises, the Danish Institute’s robot labs are in Odense. Both can trace their origins to research at the city’s university decades ago, as can a number of other robot companies and consultancies in the area. Such clusters take time to come into their own, which may be another reason why robotics research feels as though it is entering a new stage of development. There was very little academic research before there were undergraduates who had seen “Star Wars” as kids; those former fans have now had a professional lifetime to build academic research groups and spin off companies.

The biggest cluster, that around MIT, is home not only to Boston Dynamics, to date supported mostly by military R&D contracts, but also to iRobot, which has shown the way in developing a profitable service-robotics business aimed at consumers. The company was started in 1990 by Mr Brooks (now at Rethink Robotics), Colin Angle and Helen Greiner, all robotics researchers at MIT’s artificial-intelligence lab. They knew they wanted to make a business out of robotics, but not what that business should be. It was not until 2002 that they came up with the two products that have made the company’s reputation: the Packbot, which has helped soldiers deal with improvised-explosive attacks in Iraq and Afghanistan, and the Roomba, which cleans floors. The company has sold more than 8m of these; last year home robots, mostly Roombas, accounted for 88% of its sales.

The company not only makes a profit; it has also rewarded its original investors by going public in 2005, something which no other robot startup has done. However, in the past few years a number of them have been bought up for tidy sums. In 2012 Amazon, a retailing giant, took over Kiva Systems, a company based near Boston which makes robots for warehouses. Google’s robotic acquisitions followed in late 2013. “Great news for the industry,” says Mr Angle.

Such proof of viable exit strategies should inspire further investment, which a new French fund, Robolution, hopes to tap. Its boss, Bruno Bonnell, explains that ten years ago he was unable to persuade the company he ran, Infogrames/Atari, to get into robots. He left to become the French distributor for iRobot’s Roombas. In 2006 he sold 800; last year the figure had risen to 100,000. In early 2014 he closed the Robolution fund for early investment in robot companies at €80m, admittedly most of it put in by the French government and the European Union. Pointing to a sharp increase in American venture funding for robots—he puts it at $400m last year—he is convinced a lot of talent can be got out of European labs too.

Another potential source of money may be entrepreneurs who have done well out of other technologies at a young age and for whom the science-fiction feel of robotics is a turn-on, not a danger signal. The space business points the way. Elon Musk is shaking it up, using some of his internet-derived riches to create SpaceX, a disruptively good rocket-maker; Amazon’s Jeff Bezos has a rocket company too, and Google’s Larry Page takes a keen interest in such things. Robots offer a similar attraction; witness Mr Hassan’s creation of Willow Garage with some of the money he made from Google. Or, indeed, witness Google itself. The company’s recent acquisitions in the field are being supervised by Andy Rubin, who has long been fascinated by robots. When he developed the operating system Google uses for mobile phones, which became a great success, he named it Android.

Promise in the cloud

Google’s expertise at dealing with huge amounts of data will almost certainly play a key part in its plans. By drawing on the computing power of cloud-based systems, its robots, and others, should be able to do much more than they are currently capable of. The self-driving car demonstrates the idea; it can mesh information on its whereabouts from its sensors with maps of the world held in the cloud, with various programs using the comparison to generate instructions for the cars’ motors, steering systems and so on. Ken Goldberg of the University of California, Berkeley, suggests that a similar use of “cloud robotics”—a term coined by a Google employee, James Kuffner—could make it much easier for robots to recognise objects for what they are and act accordingly.

The cloud already houses libraries and programs that can help computers work out what an image is of, and robots to work out from the shape of an object how to pick it up. The European Union’s “RoboEarth” project imagines a cloud-based system that would contain all sorts of such information in a form that robots could use, and that would let robots learn from each other, both about the world around them and about successful ways of tackling tasks in that world.

Far better computers, on board and in the cloud; good, cheap sensors; maturing industrial-academic clusters; a broadening and speeding up of the field’s research base; expanding markets; exciting hardware; and a newly encouraging investment outlook: all of these have helped stimulate interest in robotics. The collapse of another science-fiction dream at Japan’s Fukushima Dai-ichi nuclear power plant in 2011 gave it an extra push. Mr Pratt traces the genesis of the DRC to the day after a tsunami hit Fukushima, when it became clear that the robots needed for such emergencies, widely believed to exist already, were nowhere to be found. Mr Inaba at Tokyo University suggests that some day emergency robots will become mandatory at big industrial installations, just as fire extinguishers are required in offices.

And a final reason why interest in robots has taken off is that some of the machines have been doing so themselves, in a very high-profile way. The greatly expanded role of aerial drones in warfare shows what such machines can achieve—and raises both hopes and fears about what they may do next.