杏吧原创

Must-have robots come nearer with software explosion

Software is about to haul robots out of the lab and into our lives, just as it did with home computing

Little Asimo didn't look like the other children
Little Asimo didn鈥檛 look like the other children
(Image: Sinopix/Rex Features)
In search of a killer app
In search of a killer app
(Image: Aldebaran Robotics)

TED LARSON walks to the front of the crowded room and lays an ordinary-looking smartphone, encased in a black plastic cradle, flat on the table.

It鈥檚 show-and-tell at the (HBRC), the informal nerve centre of Silicon Valley鈥檚 robotics community and the heart of a technological revolution that is about to haul robots out of the lab and into our lives.

Larson strokes a finger across the phone鈥檚 touchscreen and an animated face appears. One end of the phone then rears up as the cradle splits into a pair of legs, allowing the phone-robot to wander around, cooing when people 鈥渢ickle鈥 it and taking photos of faces using its inbuilt camera. 鈥淚t鈥檚 a transformer!鈥 laughs one delighted hobbyist. 鈥淚t鈥檚 Skynet,鈥 breathes another.

They may both be right. Larson, who runs a robotic toy and gadget company called , believes roboticists like those in HBRC are on the cusp of a breakthrough. Just as software transformed computers from dull accounting tools to can鈥檛-live-without devices, it is about to do the same for robotics. HBRC is an off-shoot of Silicon Valley鈥檚 Homebrew Computer Club, which dates back to the 1970s when it counted Apple鈥檚 founders Steve Jobs and Steve Wozniak among its ranks.

鈥淚 feel like robotics today is where computing was in the 70s,鈥 says of the AI lab at Stanford University, California.

The software robot revolution would shift the way we think about robots. No longer would they be purpose-built objects with preset functions. 鈥淣ow the emphasis is more on the software,鈥 says , a roboticist at the University of Pennsylvania鈥檚 GRASP lab in Philadelphia. This could be just what is needed for robots to become the must-have devices we all rely upon.

The HBRC meeting was at of Menlo Park, California, the six-year-old company that is driving much of the software revolution. Willow Garage鈥檚 major contribution has been the PR2: wheeled robots 1.5 metres tall with boxy torsos, two long, jointed arms with pincer hands, and a pan-tilt camera for a head. They are widely regarded as the most advanced general-purpose robots ever built.

Willow Garage reasons that by providing such capable hardware, roboticists are freed up to focus on the software. 鈥淲e care about turning robots into a software problem,鈥 says CEO Steve Cousins. 鈥淪oftware is what makes devices hum.鈥

鈥淲e care about turning robots into a software problem 鈥 software is what makes devices hum鈥

At the GRASP lab, researchers have that allows the PR2 to identify and read signs, as well as predict the movement of people and navigate efficiently around them. And at the Georgia Institute of Technology in Atlanta, a PR2 has been used to help a quadriplegic man move objects around his house, control a mouse and even scratch his face.

PR2s are rare and expensive, however, costing $400,000. 鈥淧R2 is like a mainframe,鈥 says HBRC president Wayne Gramlich, likening them to the large computers that predated the PC, one of which would be shared by a whole university department. 鈥淓verybody wants a mainframe but can鈥檛 afford one.鈥

Willow Garage has also developed ROS, a robot operating system that the company released in 2009. As in a PC, the operating system manages the robot鈥檚 hardware and provides an interface between it and any new software that is written for it.

This makes life easier for robotics researchers, says Ng. It is now possible for someone who wants to focus on human-robot interaction, say, to concentrate only on writing code that controls a robot鈥檚 gestures, and to use the standard code for all the other robot tasks, such as navigation. And because ROS is open source, the navigation part gets better and better, as researchers who are interested in navigation improve it before adding their modifications to the central software bank.

ROS has also accelerated innovation by stopping roboticsts from repeating each other鈥檚 work. In the past, there was no standard operating system so many researchers had to write their code from scratch. Now, with so many people using ROS, it鈥檚 easy to share applications.

Still, ROS isn鈥檛 quite as easy to write for as the major computer operating systems. And the robots that run ROS are still quite expensive. The ROS-compatible Nao (see photo), made by in Paris, France, for example, might be considered cheap by researchers but still costs around $15,000 鈥 rather more than your average PC.

That鈥檚 where Larson and his smartphone robot comes in. 鈥淲e think cellphones are going to be the new way to make the next hot robot,鈥 he says. The creation he debuted at the HBRC meeting is called a Phonedox and runs on Google鈥檚 Android operating system.

鈥淐ellphones are going to be the way to make the next hot robot 鈥 their sensors make them promising鈥

He points out that phones are now 鈥減owerful computers鈥 that come with an operating system and sensors that can be used to a robot鈥檚 advantage, including a camera and gyroscopes. His money is on telepresence as the killer app: video chat software could turn a cellphone or tablet on a chassis into a cheap remote-controlled physical avatar of yourself that attends meetings.

Cynthia Breazeal and colleagues at the Massachusetts Institute of Technology, who study the social and emotional aspects of robotics, are also banking on cellphones. They recently created , an inanimate fluffy orange dragon that turns into a robot when a cellphone is inserted (pictured above). Like the Phonedox, the screen becomes the face and the Android operating system its brain. What鈥檚 more, as DragonBot is constantly connected to the internet via its cellphone connection, its memories and 鈥減ersonality鈥 can be stored in the cloud, on a remote server.

As well as freeing up a robot鈥檚 limited computing resources, the cloud also allows them to communicate with one another, making each individual more intelligent.

This is how Carlos Asmat sees it. He鈥檚 a robot app developer in Montreal, Canada, who is part of the team behind , which launched last month. Described by Asmat as a 鈥淔acebook for robots鈥, the site allows registered robots to post updates such as temperature, whether it is stuck behind the couch, or noting the number of people in the house. Other robots could read these updates and then use the information to guide their own actions. For example, a vacuum cleaner robot might deduce there had been a party and clean.

Another way to make robots more useful 鈥 also reminiscent of personal computers 鈥 is to let consumers choose what software runs on them. That鈥檚 the idea behind the , created by Elad Inbar, based in San Francisco, California 鈥 a deliberate echo of the app stores that currently exist for smartphones.

At the moment, the site is open for developers to upload the apps they have created and so far more than 200 have been uploaded, says Inbar. These include an app that turns the Nao into a storyteller for children and one that can enable the Nao to recognise when a pet鈥檚 dish is empty and refill it with food.

When 500 robot apps have been uploaded 鈥 and approved 鈥 then the site will be open to consumers. Inbar believes that apps are the key to turning robots into useful devices because people can choose for themselves what they do with them, and because they allow the same robot to be used for a number of different things.

Such a robot revolution may be even more powerful than that in personal computing because robots elicit different emotions. 鈥淵our computer, you don鈥檛 get attached to it, don鈥檛 feel compelled to use it or care for it,鈥 says Asmat. 鈥淏y contrast, people get attached to their robot, people project their emotions on to their robot.鈥

Topics: Robots