University professor first in line to test transformative robot “Cassie”

Tuesday, February 21, 2017 - 2:00pm

More like this

If University of Michigan College of Engineering researchers continue to prototype their innovative work, you might soon be receiving packages from someone other than a mailman — instead, a robotic bird. 

The University will be the first institution to program and test “Cassie,” a robot with the mobility of a bird and the potential to transform efficient energy usage. The robot, created by the Agility Robotics program at Oregon State University, will be arriving at the lab belonging to Jessy Grizzle, professor of electrical engineering and computer science. Grizzle has also worked on robot bipeds — machines that use two legs to walk — similar to Cassie, which takes the form of a large bird. Grizzle is well-known throughout the nation for his work on bipeds and has collaborated with Oregon State University before — which is why Cassie will be tested by his lab. 

Grizzle said Cassie’s inventors did not plan on its aviary appearance, but tweaked their design when mathematics proved that the robot would perform better with this form.

“Agility Robotics started with the design of the Atrias-series robot that we have in our lab, and then (we) used the mathematics to arrive at the robot's design,” he said. “Hence, Cassie ended up looking like a bird as a result of math, but wasn't designed that way in the sense of bio-inspired.”

Grizzle’s lab will be programming and testing the algorithm to control Cassie, in the hopes they can steer the robot toward delivery-based functions. Xingye Da, an Engineering graduate student who works in Grizzle’s lab, said this functionality was designed to maneuver through obstacles.

“Cassie was designed for carrying and delivering packages,” he said. “In delivering a package to the front of your home, the robot will probably have to climb up the stairs and go through other obstacles.”

Da added Cassie would be productive in delivering objects because of its studier structure, which can mitigate the amount of damage coming from collisions and falls.

Cassie’s functions surpass those of older robots: Grizzle’s current robot, MARLO, is another biped that “walks,” but it requires human assistance to stand up and maneuver its path. Moreover, Cassie is designed to be able to reorient its “feet,” which makes it able to turn in its course of action — a function MARLO lacks.

“The models for MARLO and Cassie were designed for outdoor walking, so the biggest challenge for MARLO is that it’s very heavy and when we take it to the place where no mobile gantry can work, like the Michigan Wave Field or the woods ... we have to carry it on a boom to support its weight in case it falls,” Grizzle said. “Cassie does not need this and is more durable.”

Along with this engineered mobility, Cassie is set to transform energy efficiency in robots. Da explained robots, compared to humans, are very inefficient in terms of expending energy.

“Compared to humans’ energy efficiency, MARLO is five times less efficient than humans whereas other robots are 100 times less efficient,” he said. “I expect Cassie will be even more efficient than MARLO because of its lighter weight.”

Da expects this energy efficiency to be a main feature of Cassie. He explained per its design for the outdoors, conserving energy will be a necessity for carrying out missions.

“If you think about doing outdoor exploring or rescue, the biggest concern is the battery life,” he said. “If you carry more power, of course you’ll end up weighing heavier, so there’s a tradeoff. But Cassie was designed to counter this.”

Cassie, along with other bipedal robots, was designed to harness a variety of human mobility. Dmitry Berenson, assistant professor of electrical engineering and computer science, predicted future robots will integrate more humanoid-type parts.

“Cassie is a pair of legs, but to do useful things you would need some way to interact with the world (with arms),” Berenson said. “It's good to develop arm and leg capabilities independently so we can focus on the core problems of each. However, we're moving toward a point where both technologies are maturing and need to be integrated into a full humanoid system.”