Alive, autonomous and well

As robotics technology increasingly moves from the theoretical to practical, clients are still coming to CMU’s National Robotics Engineering Center—just not with ‘easy projects’

Herman Herman (CS 1993, 1996)

In the strawberry-growing regions of California, nearly every farm displays a “help wanted” sign during harvest season. Over the years, the labor shortage across the state’s 40,000 acres of strawberry farms has been growing progressively worse. One farm would raise pay to attract laborers from the farm next door, and another would then follow, increasing labor costs in an ever-escalating wage war. Farms started making greater use of the federal H-2A agricultural guest worker program to import seasonal help from Mexico, but the process was costly and paperwork intensive.

Then, in 2009, a farm owner had an idea: Task a robotics firm with inventing a technology to reduce the labor needed to sort plants for the strawberry nursery. The development cost would be stiff, so the owner of that farm, Lassen Canyon Nursery, convinced some colleagues to form an LLC to fund a future for strawberry processing. They took the project 2,500 miles east, to CMU’s National Robotics Engineering Center.

“We didn’t even have a mature technology for them [at the time],” recalls Jeff Legault, NREC’s director of strategic business development. “They had some seed money and we agreed to take a look at their problems.”

Five years later, NREC researchers had a prototype: a conveyer belt monitored by robot eyes that can classify strawberry plants as “good quality,” “bad quality” and “questionable” based on appearance. The device uses jets of air to separate the three kinds of strawberry plants, leaving the “questionable” ones for human inspection. By eliminating this one task, the machine could greatly reduce costs for large farms.

This is typical of NREC’s method: Someone has a complex problem and the engineers craft a solution using robotics. Whereas NREC’s projects once came almost entirely from U.S. government contracts—particularly branches of the military—increasingly, NREC’s clients are from the business world. Herman Herman (CS 1993, 1996), who was named director of the nearly 20-year-old institution in February 2015, says NREC’s work is now roughly “50/50” split between the public and private sector.

“If this was a place where everyone was working on a Ph.D., the focus would be solving (purely) technical problems,” Herman says. Instead, he says, NREC is unique in academia in that it tackles real-world challenges in ways that have immediate, practical benefits. “We have to not just solve technical problems, but resolve those problems in a cost-efficient manner,” Herman says.

Another trend is that clients are coming to NREC, as opposed to NREC searching them out, Legault says. Part of the reason is that robotics technology is becoming more affordable, and thus more mainstream. “In the past, there often was viable technology we could develop, but it would have been too expensive to implement commercially,” he says. “That’s now changing.”

NREC is located about three miles from CMU’s Pittsburgh campus in a warehouse-like building in the Lawrenceville neighborhood. About 100 people work at the center at any given time, including a mix of full-time NREC staffers, school faculty and a handful of students. 

On the inside, the facility looks like something from a comic book movie—maybe a secret subsidiary of one of Tony Stark’s companies. The open floor contains a prototype Humvee programmed to create digital maps and change detection as it roams—sort of a version of the cars that take Google Street View pictures, but for dangerous or war-torn areas. Nearby, there’s a self-driving all-terrain vehicle called “REC-Tamer,” which can be lowered from a helicopter into a disaster zone. And there’s CHIMP, a versatile robot with orangutan-like arms designed for search and rescue missions. Built at NREC, CHIMP won $500,000 in the June 2015 DARPA Robotics Challenge, a two-day event that tested 24 of the world’s most advanced robots against one another in a series of disaster-relief tasks. A looping video on display shows CHIMP picking up and using power tools to enter a mockup of a disaster area.

Whether they’re from the private sector or the military, NREC clients these days want robots that are autonomous—able to act and maneuver with little or no human input. REC-Tamer, for instance, was developed for the U.S. Defense Department by NREC in cooperation with Sikorsky Aircraft. In a demonstration Oct. 27 in West Palm Beach, Fla., a modified Black Hawk helicopter autonomously flew itself to a remote location and lowered the REC-Tamer to the ground, where it autonomously navigated over six miles of terrain, surveying environmental conditions as it went. The project also won the “Distinguished Engineering Project Achievement Award” from The Engineers’ Council.

The project was in part spurred by the disaster at the Fukushima nuclear reactor in Japan, Herman says. “After Fukushima, the U.S. Army realized it didn’t have an autonomous vehicle that could go into such an environment to survey damages or contamination. People often forget that the army is in charge of a lot of rescue and humanitarian work.”

The NREC building also has a prototype autonomous tractor, commissioned by Deere & Company, which can perform agricultural tasks, such as spraying or mowing, on its own. The prototype has accumulated thousands of miles of autonomous operation during extensive testing in Florida.  There’s even a boat inside the Lawrenceville facility, propped up on a hitch trailer, which NREC engineers are enhancing to be able to navigate by itself to and from docks, picking up and discharging passengers—a self-driving water taxi. To facilitate all the fabrication of the robots, the building includes an in-house full machine shop with 3-D prototyping capability and a temperature-controlled test chamber where robots meant to function in severe weather conditions, such as the arctic or the desert, can be tested.

It’s hard to generalize about the process of commercializing technology developed at NREC, because it varies greatly, depending on the scope of the work and the market for which it’s intended—Herman says it can take anywhere from one to five years and cost between ”hundred of thousands” and “tens of millions.” But once NREC develops and extensively tests a solution for a client, those clients can have their finished product fabricated or commercialized by one of several different specialty manufacturing companies—many of which are spinoffs from CMU, founded or run by alumni or former NREC employees, and which often are located in the Pittsburgh region. CMU retains a stake in technology created at the university, which creates revenue streams to support its programs.

To facilitate the commercialization of its technology, part of the NREC facility is dedicated for incubating spinoff companies and related activities. One entity that is currently a resident at NREC is The Robotics Hub, an independent startup accelerator meant to fund robotics high-tech companies in Pittsburgh, including additional NREC spinoffs. The Robotics Hub was started with a seed funding from GE Ventures, an investment firm associated with General Electric, to pursue various commercial applications for robotics technology created at CMU. 

Last spring, NREC made headlines, but not for its world-changing research. In February 2015, the ride-sharing company Uber established a Pittsburgh research lab to develop driverless vehicles, and hired about 40 NREC employees. Media reports used words like “gutted” and “completely raided” to describe the hiring process; Herman says it was nothing of the sort, and most projects continue as usual at NREC, where staffing levels have been “healthy and growing again,” he says.

“I think what is lost in the media reports is the reason why Uber came to Pittsburgh to establish their research lab,” Herman says. It’s because, in his words, Pittsburgh has an “enviable wealth of talent” with top researchers being educated and employed at centers such as NREC. “I’ve always said, if you work at NREC for three or four years, you can get a job anywhere,” Herman says, adding that it’s never been unusual for NREC staffers to leave for private enterprise. In fact, the center prides itself on the number of its alumni who have started robotics spin-off companies or taken important positions in developing autonomous robotics technology at companies such as Google.

“We’re accustomed to a natural flow of technical and research talent back and forth between academia and industry,” says Andrew Moore, dean of SCS and a professor of robotics and computer science. Moore himself took a leave of absence from CMU in 2006 to launch Google’s Pittsburgh office. Although four Robotics Institute faculty are now on leave working at Uber, Moore says, CMU has hired four new faculty in robotics as well as another 13 in machine learning, systems and algorithms.

In addition, NREC has hired 10 new technical staff members in the past six months, Herman says, and plans to hire another five to 10 in the coming months to augment its existing staff. “Our staffing has always reflected the number of projects we’ve had, and we have the advantage of being part of the Robotics Institute, which has a deep pool of technical talent to draw upon in term of faculty, staffs and students.” he says.

If the future research direction of NREC isn’t easily predictable, that’s the nature of working on advanced technology, Herman says. The autonomous vehicle technology that currently underpins many of NREC’s current projects, for instance, didn’t even exist in a cost-feasible form six years ago. What will NREC be working on six years from today? No one is certain, but NREC researchers say they’re sure they’ll be called upon. The “sky is the limit” in terms of new robotics applications, Herman says. “We’re encouraging companies to dream big, and we’ll make it happen for them,” he says.

Most NREC clients—such as mining companies, aircraft manufacturers and the U.S. military—already have top-notch engineers working for them, Legault says. They only outsource the projects that are the most perplexing to their qualified staff. “They come to us with difficult problems that require a great level of expertise to come up with the unique solution,” he says. “People don’t come to us with the easy projects.”

Nick Keppler is a Pittsburgh-based freelance writer who frequently contributes to magazines and websites such as Vice, Slate, Nerve and The Village Voice.

For More Information: 

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