Jonathan Hurst automates human movement.
In a light-filled, L-shaped office overlooking the fast-moving Willamette River, a leggy robot is hanging from a long wire affixed to the ceiling. Cassie, as the bipedal robot is named, completely dominates a viewer’s attention in this open office space.
Workstations, a conference table at one end of the room and the half dozen people quietly going about their tasks seem entirely secondary.
Perhaps that’s fitting, as Cassie is the culmination of years of work by Jonathan Hurst and, more recently, his group of collaborators at Agility Robotics in Albany.
With its anthropomorphic legs and a pelvis-shaped control brain situated between those legs, Cassie is the first commercially available automaton designed to move as humans and animals do. Hurst, 38, has worked his entire career on what roboticists call “legged locomotion.”
To put it in nonrobotic lay terms, the startup’s chief technology officer is engaged in the art and science of making machines walk, run, navigate stairs and anything else we humans do to get around.
There are myriad challenges still to overcome. But to see Cassie walk along a footpath or navigate a grassy field makes one imagine that the future — with machines mimicking our own mobility and performing tasks as well as or better than us — is already here.
Paradoxically, both things are simultaneously true. Making robots as mobile as we are is laborious work, work that the tall, rangy and bespectacled Hurst has pursued since seventh grade, when he began building robots for Science Olympiads.
“Oh, Legos,” he says when asked about his favorite childhood toy. “But not just the blocks. The gears and the motors and everything in order to move.”
Hurst never admired his models once built. He immediately disassembled them. “Build it once, take it apart, build whatever we can now build because we have those parts.”
Not much has changed in Hurst’s approach to moving robots. However, we seem also to be entering a version of Moore’s law for robots, in which innovations are accelerating the pace of advances. Cassie is a significant step on that road.
“It’s the first time that any dynamically capable robot is for sale,” Hurst says.
The robot in question, it should be noted, is not named for a female, and Hurst is emphatic in pointing out that Cassie sidesteps gender issues by being, neither a “he” nor “she” but rather an “it’.” The name, he says, is inspired by cassowaries, birds native to New Guinea that are similar in size and appearance to the ostrich and the emu.
“But we did not derive the design from a cassowary. The name was derived from the way the robot looks, not the other way around.”
Efficiently powered by a six-hour battery, Cassie can stand in place without support, sit down, squat and, of course, walk. But to do something specifically useful in the human world, such as rush into a burning building and check for stranded people, much work, both in hardware and software, is yet to be done.
Hurst says Agility will first sell Cassie, for approximately $250,000, into a specific niche — i.e., other academic robot labs working on legged locomotion. But his dream is bigger.
“Our longer-term vision is about commercial applications,” Hurst says. “Legs are one of the last pieces of the puzzle for how to do package delivery. Flying drones and wheeled robots are going to have their place, but probably not in dense cities and the many places where legs are going to be a much better solution.”
Hurst envisions Cassie robots hopping off the back of FedEx trucks and stepping over curbs, skipping up the steps of a home or apartment building to leave packages on the porch while navigating past a dog or kids playing kickball.
Cassie may also do reconnaissance work for the military, collect GPS data and perform search and rescues.
A native of Fort Collins, Colorado, Hurst built on his Lego projects by pursuing a bachelor’s degree in mechanical engineering from Carnegie Mellon in Pittsburgh, where he stayed to get a master’s degree and a Ph.D. in robotics.
He came to Oregon State University in 2008 as the institution’s first robotics faculty member, working as an assistant professor in the School of Mechanical, Industrial and Manufacturing Engineering. OSU gave Hurst generous physical space for a Dynamic Robotics Laboratory, and Hurst helped grow the robotics program at a quick pace.
Over time, with the aid of grants from DARPA (Defense Advanced Research Projects Agency), Hurst and a team of student roboticists constructed ATRIAS, a bipedal robot that demonstrates Hurst’s preferred method for how machines mimic human and animal walking.
ATRIAS confirmed Hurst’s confidence in using a “spring-mass” model (think of how a pogo stick works) of mobility. However, he says ATRIAS, built by a small army of undergraduates, was “chock-full of design mistakes,” not too reliable and unable to stand in place or steer itself.
Cassie is carefully designed, can stand on its own and — importantly — has feet.
Hurst, who is married with two young daughters, was open with OSU from the start about his plans to eventually go entrepreneurial.
By the end of 2015, Hurst, who remains an OSU assistant professor, had partnered with another Carnegie alum, Damion Shelton, and Mikhail Jones, an OSU master’s student, to create Cassie. Shelton provides the business savvy; Jones is Agility’s chief engineer.
Agility does have a competitor, Boston Dynamics, an MIT spin-off that has pursued both legged and wheeled robots. What sets them apart from Boston Dynamics’ approach to robot development is Agility’s willingness to share some of its research, and its desire for Cassie to become a robot “standard for academic research around the world,” Hurst says.
“The fact that this stuff Agility is doing is open platform — well, that’s how to spread the wealth of knowledge,” says Jessy Grizzle, a colleague of Hurst’s and a mechanical and electrical engineering professor at the University of Michigan in Ann Arbor. “This is unique. Comparatively, Boston Dynamics is a black hole.”
A subsidiary of Google parent Alphabet, Boston Dynamics was offered for sale last year but hasn’t yet sold.
Hurst says Agility retains enough of the “secret sauce” about Cassie’s functionality, through patents, to stay ahead of competitive efforts. In an initial funding round last year, Coal Hill Ventures of Pittsburgh, Oregon Nanoscience and Microtechnologies Institute, and the Willamette Angel Conference provided seed money.
Hurst says Coal Hill, which specializes in venture capital for robotics start-ups, was eager to come onboard “even before Cassie existed as more than a finished CAD drawing.”
Within the next month, Agility will start working on arm-like appendages for Cassie, which will help the robot maintain balance in tricky situations, as well as a suite of sensors to help with obstacles such as stairs.
Agility has also presold three robots, to Grizzle and two other domestic robotics’ labs. The sales mean moving into a very different realm, with business issues like manufacturing and customer service to attend to. Hurst hopes the academic market will be forgiving as Cassie encounters the real world, as the robot is “by no means a Toyota, in the sense that it’s not yet a reliable device with people using it and abusing it daily.”
Negotiating the inevitable tendency for others to anthropomorphize the robot is another challenge. On the one hand, the gender-neutral Cassie will interact with humans, so it must be likeable.
“We’re not trying to create a personality yet, although that’s going to be important when you are trying to do things like [have the robot] deliver packages,” Hurst says.
On the other hand, Cassie is the result of engineering decisions. Trying to model its mobility too closely on animal morphology, Hurst says, would give you many features that are “evolutionary baggage” rather than tied to walking efficiency. Engineers also want to steer clear of the “uncanny valley” in which a robot hews too closely to human or animal features but is perceived as somehow “off” and is reacted to with revulsion.
“You want to be very careful about how people perceive the machine,” Hurst says. “It can’t be threatening.”
Hurst, who speaks with the calm self-assurance of a scientist-technologist, rejects the notion that robots like Cassie are coming to take our human jobs. The time frame is still in years and even decades, and new jobs for humans will evolve.
Yet he does think roboticists should get involved with the ethical questions that will bubble up as robots begin to walk around amongst us.
“It’s clear to me that people who understand the technology are going to need to participate in creating the policies,” he says.
Thus far, the most significant effort to deal with the ethical issues is going on not in the United States but through the European Union’s effort to create a voluntary ethical conduct code for robot designers.
To the idea that maybe there are mysteries of human and animal locomotion that shouldn’t be re-created in robots, Hurst just says, “Bah. We’re going to figure out how feet work; we’re going to figure out the actuation of arms; our plan for stairs is coming up; we’re going to work on all that’s needed to have robots get around in the world.”
He adds: “We just have to make good decisions and do good things, useful things, with what we know.”