If you’re searching for the world’s best robotics research facility, you won’t find it in Silicon Valley, Boston, or Seattle. You’ll also be on the wrong path if you look anywhere inside Japan, China, Taiwan, or Germany.
To find the world’s premier robotics lab, you’ll have to travel to Pittsburgh, a city better known for its once-thriving steel industry, Iron City Beer, and legendary baseball and football teams than its high-tech research.
Once in town, head for Carnegie Mellon University, and you’ll soon see more robots doing more things than just about anywhere in the world.
More than 50 full-time faculty members and 180 graduate students are currently developing systems spanning an array of fields, including space robotics, medical robotics, industrial systems, computer vision, and artificial intelligence.
“It is a big place,” said Matt Mason, CMRI’s director. “There’s more than 600 people on the payroll; we work on all kinds of robotics technologies.”
The institute is also a pacesetter in robotics education. With programs ranging from a Ph.D. in robotics to summer camps for grade-school students, CMRI is a leader in inspiring and educating the next generation of roboticists.
CMRI started small
CMRI traces its roots back to what now seems like a relatively modest corporate endowment. “Westinghouse gave us a million dollars a year — in 1979, that was considered serious money — and we were off,” Mason said.
CMRI used the funds to pioneer robotics education at the graduate and undergraduate levels. “The first Ph.D. in robotics started here; now there are Ph.D. programs elsewhere,” Mason said. “Until a few years ago, if you had a Ph.D. in robotics, you were from Carnegie Mellon.”
Yet CMRI is probably best known for developing a wide range of innovative robotic technologies. “We award degrees at all levels, but our primary mission is still research,” Mason said.
Every faculty member is responsible for his or her own research funding. “Roughly one-half comes from the Defense Department,” he added. “That’s many different agencies, including DARPA [the Defense Advanced Research Projects Agency], the Army, the Air Force and the Navy.”
The National Science Foundation (NSF) is responsible for almost 15 percent of research funding, Mason said. Industrial sponsorships are also important.
“We’re getting about 15 percent … from private industry,” he said. “That’s manufacturing, mining, agriculture, and a lot of other industries.”
Other federal agencies, including the Department of Agriculture, the Department of Energy, the Department of Homeland Security, and the National Institute of Health, are responsible for most of the other research funding.
Promise and peril for CMU
The robotics faculty is at the core of CMRI’s mission. Many are visionaries and experts in existing technologies and applications.
Illah Nourbakhsh envisions a future that’s ripe with both promise and peril. In his recent book, Robot Futures (MIT Press), the CMRI robotics professor makes the case that robots are not just astounding machines, but an entirely new species that bridge the material and digital worlds.
Robot Futures looks ahead to a not-too-distant time when robots are both ubiquitous and highly capable. Robots will serve as physical avatars, Nourbakhsh predicts, allowing people to interact simultaneously with others in a variety of locations and situations.
Robots may even enable people to assume new and different forms, and may well change perceptions of what it means to be human, Nourbakhsh stated.
Carnegie Mellon researchers are working to prepare a new generation of robot builders and users with the help of Finch, a small robot specifically designed for computer science education. Finch’s design is the result of a four-year study at CMRI’s CREATE lab. The robot is designed to introduce its users to the art of programming.
The $99 device offers support for over a dozen programming languages and environments, including several appropriate for students as young as eight years old. Finch’s on-board features include light, temperature, and obstacle sensors. It also includes accelerometers, motors, a buzzer, and a pen mount for drawing.
Finch is being marketed to educational institutions by Pittsburgh-based BirdBrain Technologies, founded in 2010 by Tom Lauwers, then a CMRI robotics doctoral student. BirdBrain’s goal, according to Lauwers, is to take educational tools research and make them useful to the educational community.
“Our aim is to create products with a low barrier to entry, so that parents, teachers, and students can focus their time on using our products in creative and fun ways,” he said.
CMRI to the rescue
Search-and-rescue robots is another area where CMRI faculty conducts leading-edge research. The Tartan Rescue Team at CMRI’s National Robotics Engineering Center (NREC) has developed the CMU Highly Intelligent Mobile Platform.
Known as CHIMP, the approximately human-size robot is designed to perform various tasks, such as climbing ladders, driving vehicles, and closing water or gas valves. CHIMP is a bit shorter and stockier than an average human: It’s 5 ft., 2 in. tall and weighs about 400 lb.
The statically-stable robot is designed to move on tank-like treads attached to each of its four limbs. When CHIMP needs to operate power tools, turn valves, or otherwise use its arms, the robot simply stands and rolls on its leg treads.
“We selected a design with a near-human form so that the robot would be able to operate in environments engineered for humans,” explained Tony Stentz, NREC director, team leader and a CMRI research professor. “Rather than design a machine that would need to balance itself as it carefully steps over debris, we are building a robot that is able to drop down on rubber tracks and roll over debris.”
On the software front, researchers are designing a system that aims to help robots discover nearby objects via computer vision and related software. By taking advantage of all of the information made available to it by vision — an object’s location, size, shape, and approximate weight — a robot can continually discover and refine its understanding of objects.
The Lifelong Robotic Object Discovery (LROD) process developed by the research team allows a two-armed, mobile robot called HERB (Home-Exploring Robot Butler) to use color video, a Kinect depth camera and non-visual information to discover more than 100 objects in a home-like laboratory, including items such as computer monitors, plants and food items.
Object recognition has long challenged computer vision researchers. A cluttered environment can make object recognition an almost impossible computational problem, said Siddhartha Srinivasa, a CMRI associate professor of robotics who jointly supervised the research with Martial Hebert, a CMRI professor of robotics.
Yet humans don’t rely on sight alone to understand objects. The researchers felt that robots, too, can be designed to possess a significant amount of “domain knowledge” about their environment that they can be used to discover objects.
Depth measurements from HERB’s Kinect sensors proved to be particularly useful, providing three-dimensional shape data that lets the robot to discriminate among objects. Location knowledge, meanwhile, helps HERB determine whether something is on a table, on the floor or in a cupboard.
The team found that adding domain knowledge to the video input almost tripled the number of objects HERB could discover and reduced computer processing time by a factor of 190.
“As the robot uses its domain knowledge, it becomes clearer what is and isn’t an object,” Srinivasa said.
Applying robotics research to devices that can help people with move more freely and with less pain is the goal of a team led by Yong-Lae Park, an assistant professor of robotics. Working with Harvard University, the University of Southern California, and the Massachusetts Institute of Technology, the researchers created a soft, wearable device that mimics the muscles, tendons and ligaments of the lower leg.
The invention could aid in the rehabilitation of patients with ankle-foot disorders such as drop foot, Park said.
The “active orthotic device” uses soft plastics and composite materials instead of a rigid exoskeleton. The soft materials, combined with pneumatic artificial muscles (PAMs), lightweight sensors, and advanced control software, make it possible for the robotic device to achieve natural motions in the ankle.
As they enhance joint function, an active orthotic device can also help re-educate the neuromuscular system, Park said.
“The limitation of a traditional exoskeleton is that it limits the natural degrees of freedom of the body,” he noted. The ankle is naturally capable of a complicated three-dimensional motion, but most rigid exoskeletons allow only a single pivot point.
Park says the new approach could eventually be used to create rehabilitative devices for other body joints, or even to create soft exoskeletons that increase the strength of the wearer.
CMU and collaboration
Mason says the CMRI faculty is key to developing new robotic capabilities while educating and encouraging students to become tomorrow’s roboticists.
“I want to brag about what a collaborative group [the faculty members] are — how open-minded, how welcoming, how easy they are to work with, and how good they are at building real systems that really work,” Mason says. “That’s why we get so much support from our sponsors.”