By Ariana Assaf, Daily Staff Reporter
Published January 15, 2014
Mythical creatures have a tendency to sparkle in sunlight, but a walking robot that exists on the University’s campus was working on something far more impressive when it ventured out into the open air.
More like this
MARLO, a bipedal robot, is helping researchers develop the science of leg locomotion for machines. Essentially, the goal of working with MARLO is to get robots to walk just as well as, if not better than, humans.
MARLO was designed and built by Jonathan Hurst, assistant engineering professor at Oregon State University, using funding provided by the Defense Advanced Research Project Agency, a program funded by the Department of Defense. MARLO is one of three robots created using Hurst’s ATRIAS principle, or “assume the robot is a sphere.” The other two robots are housed at Oregon State University and Carnegie Melon University.
For the past two years, Engineering Prof. Jessy Grizzle has led a team of University scientists on a quest to program MARLO in the best, most stable way possible. He said in the future, these robots could be used to assist in situations that might threaten human life.
Grizzle said robots like MARLO could play a critical role in recovering from natural disasters like the tsunami that hit Japan in 2011. The tsunami struck the Fukushima nuclear power plant and caused enormous damage and subsequent radiation leaks that proved too dangerous for humans to try to fix.
“We want machines to go into situations that are too dangerous for humans, or more dangerous than we’d want to be sending humans in to,” Grizzle said.
The team began their work with MABEL — an earlier and slightly more primitive version of MARLO — whose hips only moved forwards and backwards. This version of the robot could not provide its own lateral stabilization and was confined to walking or running in circles inside the lab attached to a boom.
MARLO, however, can walk in all directions attached only to a safety cable that prevents it from toppling over and breaking when it loses balance. This improved ability to balance allowed the team to take MARLO for a test run outside the lab. On one chilly day in early December, MARLO finally saw the light of day.
“We’d done the length of the lab 25 or 30 times, but these robots are eventually supposed to be able to go outside and handle real walking situations,” Grizzle said. “One Saturday morning we took a shot at it, and it almost worked.”
MARLO walked around for about two hours until a knee broke. It took roughly an hour of sanding and gluing to put the pieces back together, and another day for the glue to dry.
“We’re still rookies at working on this robot, but we’ll lick that problem soon,” Grizzle said.
Engineering graduate student Brian Buss said he was attracted to the project because of the way it attacks such a complex problem, and he appreciates the potential it has to help people.
He related programming MARLO to teaching a child how to walk.
“A child is born with limbs and muscles, but can’t walk until it learns how,” he said. “The programming part we’ve done consists of figuring out what to do with six motors and sensors that measures angles if the robot is leaning over.”
Buss said balance is a trickier thing to accomplish than most people realize. It’s not as simple as some may think; when compared to MARLO, humans have the advantage of nerves that detect pressure and tipping angles, eyes to see where they’re going and flexible ankles that can correct balance issues.
The team plans to install high-performance internal measuring units on MARLO to give the robot an even better sense of balance. Currently, devices at the ankles measure forces on the ground while others calculate the angle of the body. However, these devices aren’t foolproof and MARLO still trips over itself from time to time.