Hollywood has come a long way since R2-D2 rolled about in Star Wars (1977) with the turn-on-a-dime sophistication of a self-propelled lawn mower. Its companion C-3PO wasn’t much more advanced. It moved with a mechanical grace reminiscent of the Tin Man in The Wizard of Oz.
Fast forward to the Isaac Asimov inspired sci-fi thriller I, Robot (2004) in which machines that look eerily like humans leap with Spiderman-like agility. And to the more laid-back robotic healthcare assistant in Robot and Frank (2012), which walks like a human as an accomplice to an elderly criminal.
Real robots are far from being able to move efficiently with dynamic forces similar to those exerted by legs, feet, tendons and muscles. But engineers at Oregon State University are making strides with a two-legged innovation dubbed ATRIAS. Designed and built in Corvallis, ATRIAS and its counterparts at the University of Michigan and Carnegie Mellon University in Pittsburgh are helping engineers to delve into the complexities of animal locomotion. Equipped with such capabilities, robots will serve a variety of practical needs.
The day is coming when these machines will walk up and down stairs, over rough terrain or on a crowded sidewalk with a safe, steady, reliable gait. In our homes, they will perform routine tasks that enhance the quality of life for people in need of assistance. They will do things that humans take for granted — grasp and turn a doorknob, open curtains and wash dishes. In more extreme circumstances, they will assist us in emergencies by entering damaged buildings, putting out fires and withstanding radiation or toxic fumes that would kill a human in seconds.
“It’s completely inevitable,” says Jonathan Hurst, a leader in Oregon State’s growing robotics program. “It’s possible to have robots that walk and run and manipulate things and do all the physical interactions that people can do. We have physical proof. I can do it. Why not a robot? Where will it happen first? That’s the question.”
A Natural Step
It will happen in Oregon if Hurst and his colleagues are successful. Hurst was Oregon State’s first “roboticist” when he arrived in 2008, lured by the efforts of OSU professors Kagan Tumer, a specialist in multi-agent control systems, and Belinda Batten, then head of the School of Mechanical, Industrial and Manufacturing Engineering, or MIME. Now Hurst is one of almost a dozen OSU researchers who focus on the challenges of building practical machines that exhibit the three primary qualities of any robot: They sense their surroundings, analyze their options and take action.
To support their quest, they have raised more than $24 million from the National Science Foundation and other agencies. The College of Engineering is raising funds to renovate historic Graf Hall on the Corvallis campus to accommodate the program.
In a basement lab equipped with a circular test track, Hurst and a team of undergraduate and graduate students designed ATRIAS to literally have springs in its step. These thin flat fiberglass plates, which have the physical properties of a diving board, are part of a system that imparts lively but controlled energy to the process of locomotion. “This is the first robot in history to demonstrate this kind of natural walking gait, where quantitatively we can say that center of mass motion, the ground reaction forces and the physics of the system match what animals are doing,” explains Hurst.
Hurst had two local shops — King Machine and Ram-Z — build parts for ATRIAS. And in addition to the robot used for research at OSU, he shipped models to his alma mater, Carnegie Mellon, and to a lab run by his long-time colleague Jessy Grizzle at the University of Michigan.
As a graduate student, Hurst worked with Grizzle to design and build ATRIAS’ predecessor known as MABEL. In 2012, they were recognized among 10 “Breakthrough Innovators” by Popular Mechanics. Since then, MABEL has been retired as a research robot and performs for the public in a traveling educational exhibit, now at the Field Museum of Natural History in Chicago.
Learning to Drive
Christian Hubicki, a Ph.D. student in Hurst’s lab, works on the software that controls ATRIAS. “This robot is like a high-end sports car. It’s difficult to control, and it takes a pro to drive it. We have to rise to be that pro,” he says. “There’s no remote control. We’re teaching this machine to walk, and we have to invent things all the time.”
The learning goes both ways. Hubicki and his peers perform computer simulations to test their ideas before they put them into practice on ATRIAS, but surprises happen. For example, to test software designed to enable the robot to walk through soft terrain, Hubicki once put a bucket of gravel on the lab’s test track. His simulations indicated that the machine would hop once or twice before losing stability. Instead, after it hopped into the gravel, the machine jumped 12 more times and then leapt out of the bucket onto the solid surface. Hubicki was shocked. “The controller (software) was a lot more stable than we predicted,” he says.
Such experiences underscore the difficulty of trying to mimic nature. “Locomotion is simple and at the same time incredibly complicated,” says Hurst. “Animals can do amazing feats of agility and stability running blindfolded in the dark over relatively uneven terrain. We’re trying to solve the problem of how legged locomotion works. If we can understand it, we can replicate it.”
For inspiration, Hurst and his team look to animals that predate humans: birds. Some species in this ancient branch of life may appear ungainly on land, but birds have successfully adapted walking and running to a wide range of body sizes. Ostriches have been clocked at more than 40 miles per hour. A guinea hen can adjust to obstacles on the fly and maintain a stable running gait.
It’s no accident that ATRIAS walks more like a bird than like the robots in Star Wars. In 1977, R2-D2 and 3-CPO seemed futuristic. Now they look quaint. In turn, ATRIAS may be the forerunner of machines that even Hollywood can’t quite imagine.
See an Oct. 29 OSU news release on birds as models of locomotion.