SAN JOSE, Calif. — Rosie the Robot could finally be coming to your home.
Willow Garage, a unique startup in Menlo Park, Calif., has designed a robot called the PR2 that bears some resemblance to the Jetsons' beloved Rosie. It's still under development, but already the PR2 can fold clothes, fetch a drink from the fridge, set the table and even bake cookies.
The robot's backers aren't ready to say just how soon the PR2 will hit the mainstream market. Right now, it costs too much, does too little and is too slow to be of interest to most consumers. But to many experts, the idea of a skilled and intelligent household robot finally seems to be drawing near.
"The technology is much closer than most people think," said Andrew Ng, an associate professor of computer science at Stanford University. "We're not yet there, but I think that in less than a decade the technology will exist to have a useful household robot."
Willow Garage, founded four years ago by Scott Hassan, a software developer who previously created the company that became Yahoo Groups, is a for-profit institution with the express goal speeding up development of such robots. Its investors, who, other than Hassan, Willow Garage declines to reveal, evaluate the company on how well it's succeeding at its mission.
Instead of tallying profits, they look at things such as how quickly developers are adding code to ROS, the operating system underneath the PR2; how many developers and institutions are using that code in their robots and whether the work done with the robots or ROS is winning industry awards.
"We're trying to build a personal robotics industry," said Steve Cousins, Willow Garage's CEO. "We want to serve as a catalyst."
Building on open source
Willow Garage operates on the premise that a primary impediment to progress in robotics has been the lack of standards. Each time a company or research institution has had an idea for a new robotic feature or task, it usually has started from scratch, building its own unique hardware and software to test it out. Because the hardware and software weren't compatible with those developed by other researchers, experiments generally couldn't be duplicated and the scientists had a difficult, if not impossible, time building on work done by others.
Willow Garage's answer was to develop a robot with plenty of technical capabilities built on top of an open-source operating system that it would distribute widely within the robotics research community. Last year, the company gave away 11 PR2s to top university and corporate-based robotics labs. It's since sold 13 other PR2s to similar organizations.
The assumption is that if robotics researchers at different institutions all use the same basic platform, they could stop reinventing the wheel and share results, code and insights. The idea was to turn the robot into something like a PC or smartphone, where people could start focusing on building programs or apps, rather than the robot itself.
For work and play
So far, those robotic apps have been impressive. A team at the University of California-Berkeley has programmed their PR2 to sort and fold laundry. A team at Willow Garage programmed a PR2 to play pool, sinking shot after shot. And a team at the Massachusetts Institute of Technology programmed their PR2 to mix up some dough from raw ingredients and bake cookies in a toaster oven.
But researchers are taking the underlying code well beyond the PR2. There are now some 3,100 "packages" of code that have been added to ROS, up from fewer than 1,300 a year ago, according to Willow Garage. And the vast majority of those packages were developed not for the PR2 but for robots such as four-rotor flying vehicles and autonomous cars.
Many companies that in the past dismissed research on robotics as too expensive are now taking a look at what can be done with ROS and inexpensive, off-the-shelf robotics components, said Hyoun Park, a research analyst at the Aberdeen Group. Such experimentation "could really change the way we think of robotics," said Park.
Analysts such as Park see a wide variety of potential applications for self-guided customizable robots such as the PR2 and its potential successors. A real-life Rosie is an obvious one, but may still be a ways off. In the nearer term, personal robots may take on roles in small businesses and warehouses doing repetitive tasks or as mobile security guards.
Help for the disabled
One area that may have a lot of potential for PR2's successors is in-home health care for the elderly or the disabled, said Richard Doherty, research director with the Envisioneering Group. With an aging population in the United States, Western Europe and Japan, there's a growing need for nurses who can monitor and assist patients. But there's a shortage of visiting home nurses, Doherty said.
But before robots become widely employed, their cost must come down. The original, two-armed PR2 retails at $400,000. Willow Garage has attempted to reduce the price by offering a model with just one arm, but that model still costs $285,000. The company offers both robots at a discount to researchers who contribute code to ROS, but even then the lowest priced PR2 costs about $200,000.
Even at a reduced price, the robots will have to be a lot more capable than they are now before they become a mainstream product. Even at the tasks they've started to master, such as folding towels or fetching a drink, robots take far longer than humans. And often their performance is inconsistent at best.
They need to get better
At UC-Berkeley, for instance, researchers have tried to have a PR2 sort through and fold a pile of laundry. But even if the number of different types of laundry articles is kept to just five or 10, the PR2 is successful at identifying them no more than 80 percent of the time, said Pieter Abbeel, a professor in the electrical engineering and computer sciences department. And if the variety increases, the robot has an even more difficult time.
"It's still a big challenge," Abbeel said.
Robots still have a very difficult time doing any number of everyday tasks, such as setting the table or putting objects away. That's because robots don't see things the way you and I do, experts say. Instead, they see "a lot of numbers," said Ng.
Still, in spite of all the challenges, personal household robots are moving from science fiction to reality.
"I feel like we're on the cusp of a revolution, Ng said.