Walking is the most natural way for humans to get around.
But for a robot? Not so much.
This is a problem Jonathan Hurst is trying to solve at Oregon State University.
Hurst is an associate professor of mechanical engineering at OSU. In a large open laboratory, Hurst and his students work every day, testing theories, developing equations, all with the same goal: get robots out into the real world.
“We want robots to be working with us,” he said. “We want robots in our homes, we want robots out and about doing package delivery, we want robots in the dangerous situation where people don’t want to go.”
Before Hurst’s arrival in 2008, OSU’s College of Engineering didn’t have a robotics program. Hurst was hired as the university’s first robotics faculty member.
“The leadership were very open," Hurst said. He was given wide latitude to build up robotics at OSU.
In 10 years, OSU robotics went from fledgling project to one of the nation’s top programs.
“I think the push towards creating things that have an immediate impact in the real world is a cool thing,” said Ph.D. student Kevin Green.
Green is one of a handful of graduate and undergraduate students working with Hurst in studying the dynamics of two-legged movement and how to apply them to robots.
They turned to the movement of two-legged animals (specifically birds) to understand how those dynamics work.
“We want to understand what animals are doing that makes them so effective at running over terrain and then figure out how to reproduce that performance in a robot,” he said.
In an experiment, Hurst and his collaborators watched a guinea fowl run multiple times across a flat platform covered in tissue paper.
Then they removed a piece of the platform and watched the bird try to navigate the unexpected drop.
Hurst discovered that even though the bird didn’t see the drop, its natural running dynamics kept the bird from getting hurt.
He combined the data from that experiment with a long-standing theory of dynamics called the spring-mass model.
Picture a person jumping on a pogo stick. Energy is stored in the spring when it’s compressed. When it expands, energy is released, sending the rider up and forward.
“A simple spring-mass model like a pogo stick … can reproduce that same behavior from animal gaits,” Hurst said.
Armed with the data about dynamics and movement, Hurst and his students built the robot ATRIAS.
ATRIAS, which stands for Assume The Robot Is A Sphere, was designed to replicate as closely as possible the dynamics of how we walk.
“So we wanted all the mass concentrated at the hips and we want the legs to be as mass-less as possible, and we wanted a physical spring between the toe and the hip,” he said.
Getting ATRIAS to walk like a human wasn’t easy. It took about a year and a half of trial and error before they finally got it right.
Students even tested ATRIAS’ ability to adjust to impact by pelting it with dodgeballs.
But in the end, ATRIAS was a success.
“ATRIAS became the first machine ever to successfully reproduce the dynamics of a human walking gait,” Hurst said.
ATRIAS’ success eventually led to the creation of OSU’s latest robot, Cassie.
At first glance, Cassie resembles the bottom half of an ostrich; its knees and thighs are pointed backward while the lower legs, ankles and toes are pointed forward.
But Hurst said the design had less to do with how the bird looked and more with how the bird walked.
In fact, Cassie’s design built upon the information gathered from ATRIAS.
“Cassie has ankles, Cassie can steer," Hurst said. Its legs, unlike prior generations of robots, can swivel left and right, allowing it to turn, mid-stride.
Cassie isn’t very large compared to most robots; it only weighs 70 pounds and stands at about 4 feet in height. But what it lacks in size and weight, it makes up for in movement and efficiency.
Cassie is piloted by a remote control and can walk on even ground easily enough.
But the real magic comes when Cassie walks on uneven terrain, like grass. Once it goes from pavement to grass, Cassie can continue to walk without much assistance.
There are no outside sensors, so Cassie relies entirely on its dynamics to navigate the area without falling over.
So what are the next steps for Cassie?
For Hurst, it's getting Cassie out to as many people as possible.
His company, Agility Robotics, manufactures and sells other Cassie robots to universities for their own research purposes.
So far, six universities have purchased Cassie robots, including the University of Michigan, Caltech and the University of Pennsylvania.
“As a first step, it allows a much greater research community to really develop and move forward,” Hurst said.
It’s going to take a lot of time and work to get robots out into the real world, but it’s clear that robots like ATRIAS and Cassie are making giant steps in the right direction.
And for Hurst, the possibilities for Cassie and robots in general are endless.
“Once you start to understand something and have an application for it,” said Hurst, "you can start to really gather some major resources to develop this thing and make thousands and thousands of robots, and get it out into the world to really have an impact.”