I recently attended EmTech MIT 2011 at the MIT Media Lab in Cambridge, Mass., which is a fascinating venue for new technology ideas. One of the highlights of this conference was a set of four-minute presentations by the TR35, or the 35 people selected by Technology Review as its annual list of 35 innovators under the age of 35.
Two of these young engineers, Pieter Abbeel of the University of California at Berkeley and Brian Gerkey of Willow Garage, had some very interesting things to say about personal robotics, specifically one named the PR2.
As you might guess, programming robots is not for the faint of heart. But if you have been reading Bill Wong’s Lab Bench, you know that the task is getting somewhat easier based on open-source efforts, among other developments (see “Frameworks Make Robotics Development Easy—Or Easier, At Least” at www.electronicdesign.com).
Abbeel is breaking new ground in robotics. He calls his model apprenticeship learning for robots. Rather than writing code to get the robot to do something, humans simply demonstrate the task, and the robot observes and then mimics the behavior. Alternatively, the robot learns when it is shown pictures of the task.
The PR2 did not appear on stage, but it was the star of a short video that showed it performing common household chores. One was matching socks and putting them in pairs. Another was cleaning up a playroom full of toys. Anyone with small children who watched just this portion of the video might be willing to plunk down some serious dough for this kind of household help.
In another part of the video, the PR2 was vacuuming the floor. Most of you are familiar with iRobot’s Roomba robotic vacuum cleaner. But the PR2 is not a vacuum cleaner itself, like the Roomba. Instead, it pushes around a vacuum cleaner just like a person would.
These few operations would probably be enough to convince people that this robot would be a worthwhile investment. But the part of the video that really impressed me was when the PR2 started folding towels.
It held up the cloth from either end, raised it high to look at it, like a human might, then placed it on the table, folded it neatly, and placed it in a pile. As the PR2 continued folding items, it placed larger towels in one pile and smaller ones in another.
This seems to be a giant leap in bringing robots into the mainstream. It appears that the average person would be able to teach the PR2 how to perform tasks just by showing it what to do.
The PR2 was developed by Gerkey’s company, Willow Garage. The company works on software from the lowest level with the Robot Operating System (ROS) to the highest—and anything in between—with its collection of ROS-enabled software packages. These tools help application engineers understand camera, video, and 3D point cloud data with the OpenCV and Point Cloud Library (PCL) perception libraries.
With the PR2 (and other robots like it), the personal robot industry appears ready to take off. Willow Garage, though, sells the PR2 as a robot platform for experimentation and innovation only. And the price of the platform is steep, ranging from about $200,000 up to $400,000.
Does a robot like the PR2 have a commercial future? I’m guessing yes, so the question becomes how much the average person would pay for a personal robot versus how much a company could charge for it. Offhand, I would say that the sweet spot would be in the $2500 to $5000 range.
However, I can also see where initial pricing would be much higher, in the $15,000 to $25,000 range for a PR2-type robot that could be programmed either with ROS or some other language—for the programmers among us—or with apprenticeship learning, for the rest of us.
I’ll put a personal robot on my wishlist and hope I’ll have a chance to purchase one soon. In the meantime, you can view videos from EmTech MIT at www.technologyreview.com/emtech/11/.