The four-legged robot crouched in the lobby of the Stata Center at the Massachusetts Institute of Technology like a silvery mechanical and cat-like predator. Then, at a command from its handler, the robot sprang into the air, executed a perfect somersault, and landed on its feet.
It was a pretty good trick to kick off a three-day celebration of the school’s new Schwarzman College of Computing, a $1 billion project to integrate computer science into every university department, from genetics to political science, and of course robotics.
The program featured students and faculty who showed off software to measure fetal health, blood pressure, poverty rates in Nairobi — even an improved algorithm for online shopping.
And that acrobatic robot certainly raised the wow factor.
The machine was developed in the robotics lab of an associate professor of mechanical engineering, Sangbae Kim, and modeled after a real-life cheetah in the wild. Guided by a remote control, a full-size model of the robot weighs about 90 pounds, and the motors in its joints can produce as much torque as a small car. But Kim considers that beast too dangerous to let out of the lab, so his team developed a 20-pound version dubbed Mini Cheetah.
Kim and his colleague, robotics researcher Ben Katz, said they are building 10 of the smaller robots for research teams at other institutions, including one for the Internet titan Google, which is helping to fund the project.
The idea is to figure out the full range of the robot’s abilities. “You might think we’ve solved all the problems,” Kim said. “Not at all.”
The robot proved him right; after its second somersault, the Mini Cheetah toppled over. Kim wasn’t worried about its mechanical health — in a year of testing, the machine hasn’t had a breakdown. It seemed to be some sort of software bug.
So the two world-class engineers did exactly what any sixth-grader would do with a balky home PC: They turned the robot off, then on again. In about a minute Mini Cheetah was again up and strutting.
The mishap showed that robot engineers still have a lot to learn about mimicking the abilities of humans and animals — much less developing uses for them in society. The cerebrum of the brain contains about 20 billion neurons, Kim said, while the part that controls the body contains five times as many. And this part of the brain is still poorly understood, he added.
Figuring out how to simulate the behavior of living organisms in silicon is just the kind of problem the new College of Computing is intended to solve.
Launched with a $350 million gift from financier Stephen Schwarzman, the college will include 25 “bridge” faculty members, who will hold joint appointments in other academic departments, such as economics, linguistics, anthropology, and urban studies. These faculty members will guide the application of computer science to each of these fields. And that might someday lead to fruitful partnerships between brain science, computing, and robotics.
Kim warned that crashing global birth rates and aging populations will require robots capable of doing housework and caring for elderly people. Ideally, they’ll be robots with legs, well suited to moving around in our homes (even if they have the mechanical pedigree of a cheetah). And getting it done will require closer cooperation between the mechanical engineers who make robots move and the computer scientists who must teach them the right moves.