Building a Better Robot Workforce

Creating jobs for humans might just start with improving the jobs done by technology


By Meghan Holohan

A student scurries over to an orange robotic arm hovering above a checkerboard. He studies the arm, and then adjusts it slightly before returning to his computer. The arm slowly moves across the board, stops suddenly and projects a green-hued mirror image of the board.

As David Bourne watches the student calibrating the robot, he explains that the robot is estimating where a certain physical object should be placed as part of an assembly, and is using the projection to show the worker who is doing the assembling the exact location. Robots are excellent at measuring--if a worker needs put together a machine, a robot can tell her where to place a clamp, down to the smallest increment. A human can do this, too, but it takes a lot more time and work. Bourne is interested in creating robots that collaborate well with humans in a manufacturing setting. 

"How do we tell the robot what to do, and then how do we get the robot to tell the human what to do?" says Bourne, a principal system scientist at the Robotics Institute at Carnegie Mellon University.

He designs robots to custom-produce small runs of items in quantities as cheaply and effectively as those done in large-scale manufacturing. He envisions an era of "pop-up" manufacturing, where companies can quickly erect a plant where robots and humans work together anywhere they might be needed. His goal is to see manufacturing jobs that have moved to other countries return to the United States.

In June 2011, President Obama visited Carnegie Mellon's National Robotics Engineering Center, or NREC, to launch his Advanced Manufacturing Partnership, a collaborative effort between U.S. companies and universities, including CMU, that's trying to create more high-tech manufacturing jobs in the United States.   

"Innovations led by your professors and your students have created more than 300 companies and 9,000 jobs over the past 15 years -- companies like Carnegie Robotics," Obama said. "But more important than the ideas that you've incubated are what those ideas have become: They've become products made right here in America and, in many cases, sold all over the world. And that's in our blood. That's who we are. We are inventors, and we are makers, and we are doers."

Bourne isn't the only professor in the School of Computer Science working on projects that might transform manufacturing in the United States.  

In his office on the Pittsburgh campus, Matt Mason places a robotic hand, once used to assemble Sony Walkmans, on a conference table. It's a metal ball with six protruding "arms," each of which could handle one task; one could pick up a tray, for example, while another could stretch a rubber band. The ball rotated so that each hand could accomplish its goal, but these hands weren't very smart, Mason says. They could only perform a very limited set of tasks.

In the past several decades, researchers have attempted to build more versatile, "smarter" hands by focusing on creating robotic appendages that function like human hands. But human hands are very complex--there are 15 joints in the fingers alone--so creating a robot hand that works like a human one is a huge undertaking. 

Mason--director of the Robotics Institute and professor of computer science and robotics--wants to make a hand that's simply designed, much like those "grabber" extensions used to pull items off of high shelves in stores, but that's also able to function with the sensitivity of a human hand. 

Mason demonstrates, using one of those "grabbers" and a bowl of mini candy bars. He picks up a candy bar with the grabber, but the candy bar is on its side. Mason wants it to be flat. A person can tell without looking if the item is flat in her fingers, but a robotic hand does not know if the bar is flat, on its side, or askew. By adding sensors and cameras, and then using machine-learning techniques to analyze the visual and tactile data it receives, the hand can pick up the pieces and hold them correctly. 

The ability to understand the orientation of a physical object is important for assembly operations, when a worker needs to grab parts from bins and put them into place on a circuit board or frame. A worker can stick her hand into a bin of parts, select the correct one, and hold it the proper way without even looking. But robots lack that sensitivity and dexterity. It's one reason why assembly of electronic gear such as cell phones is done in developing nations, where human labor is inexpensive and plentiful.

Mason has a prototype of a more sophisticated robotic hand that he demonstrates by picking up dry-erase markers. With just three slender metal fingers, it looks like the "claw machine"--familiar from arcades and family restaurants--that was depicted in the movie "Toy Story." (In fact, Mason says that the researchers have been testing the hand with a bunch of the toy aliens from "Toy Story.") But unlike the "claw," it requires no human intervention to correctly orient the dry-erase markers. A robot like this prototype--working alongside humans--could one day help bring some electronics assembly jobs back to the United States. Robot and human employees together could make U.S. manufacturing less expensive than overseas factories that rely solely on human work. 

While Mason and Bourne are developing prototypes, John Bares, former director of NREC, is trying to turn prototypes into products. Bares is a co-founder of Carnegie Robotics. He says that professors often develop prototypes and want to commercialize them, but do not have the time or capabilities to do so, so they ship their prototype to an outside company for commercialization. Carnegie Robotics is perfectly positioned to be that other entity, he says. The company conducts its product engineering and development in Pittsburgh and tries to keep robotics fabrication in the region by encouraging local companies to bid. 

But Bares adds that he thinks more broadly, and he's happy if he can keep the work in the United States. "We have consultants and [subcontractors] across the United States, but 75 percent of the work is done [in Pittsburgh]," he says.

While he can't divulge many details, Bares notes that Carnegie Robotics is working on several defense-related projects to build robots that detect IEDs in warzones. And the company is developing sensor technology such as its new product called "EyesOn," which will allow remote operators of robots to develop "situational awareness"--an understanding of the whole environment and surroundings in any area in which they're operating a robot.

Tony Stentz, current director of NREC, calls the relationship between the center and Carnegie Robotics "quite beneficial. At NREC, for instance, we do a lot of work that is sponsored by third parties and they are quite often very interested in generating a robot and actually using that robot to solve some particular problem. If we don't have the full story--from the basic development to making multiple copies--they might not be interested in dealing with us."

Reversing a decades-long decline in U.S. manufacturing jobs won't happen overnight. But nationwide partnerships like the AMP--and local partnerships like those of CMU with Carnegie Robotics and its network of subcontractors--provides some hope that such a goal can be achieved.

Meghan Holohan is a Pittsburgh-based freelance writer whose work has appeared in PittMed, MentalFloss.com and Salon.


For More Information: 

Jason Togyer | 412-268-8721 | jt3y@cs.cmu.edu