ATRIAS is a series of prototype human-scale bipedal robots designed and built by the Oregon State University Dynamic Robotics Laboratory. They are designed to test and demonstrate theoretical concepts for efficient and agile locomotion, ultimately aiming for walking and running outside in rough terrain.
ATRIAS is designed to move like a simple “spring-mass” model, a theoretical model which is comparable to a pogo stick. This springy model can both walk and run with remarkable energy economy and in a fashion highly similar to humans and other animals. By building ATRIAS like this model, we are targeting similar performance.
ATRIAS' unusual leg design is central to its agility and efficiency. The four-bar carbon-fiber leg mechanism is very lightweight, softening each footfall instead of sending large jolts to the body. The legs are also mounted to series-elastic fiberglass springs, which act both as a suspension and a means of mechanical energy storage. These features allow ATRIAS to save energy and execute more dynamic maneuvers.
Humans are far from the only members of the animal kingdom to run on two legs. Ground-running birds, like ostriches, can run with amazing speed and maneuverability. In collaboration with Dr. Monica Daley of the Royal Veterinary College, we use the physical principles encoded in ATRIAS to discover how birds ranging from 250-lb ostriches to tiny quail run with such an agile and stable fashion.
There are three ATRIAS robots in operation around the United States. Multiplying robots means multiplying the rate of research progress. Dr. Jonathan Hurst's lab at Oregon State University, Dr. Jessy Grizzle's lab at the University of Michigan, and Dr. Hartmut Geyer's lab at Carnegie Mellon University are all working with ATRIAS prototypes.
ATRIAS will be appearing for a live demonstration at the DARPA Robotics Challenge in Pomona, CA on June 5-6, 2015. We look forward to showing you ATRIAS walking around outside, in person.
ATRIAS, OSU’s very own bipedal robot, has a big show coming up.
In just three months in Pomona California, the Defense Advanced Research Projects Agency (DARPA) will be throwing the world’s biggest technological soiree: the DARPA Robotics Challenge. The event will assemble robotics industry professionals, thousands of spectators, and of course... the most advanced robots from all around the world. ATRIAS, a human-sized two-legged robot designed and built by Oregon State’s Dynamic Robotics Laboratory, received an invitation. DARPA would like ATRIAS strut its stuff right in front of the grandstands. That’s center stage.
Some robots are coming to Pomona to compete in the headline event, a $2 million challenge prize for responding to a simulated industrial disaster. The contest: drive a tractor to the disaster site, climb over rubble, cut through a wall with an electric drill, open and walk through a series of doors, turn a pressure-relief valve, and perform a variety of other emergency-related functions.
ATRIAS is a prototype of the next generation of these disaster-response machines. The robot will showcase its own talents at the DARPA Robotics Challenge: walking and running with agility and efficiency. ATRIAS can get off the ground and jog like a human. ATRIAS also consumes only a small fraction of the power of other bipeds, allowing for much longer battery life.
This ability to maneuver quickly and efficiently is ATRIAS’ primary goal, which shapes its unusual mechanical design. Robots don’t typically have such spindly legs, such tiny feet, or big springs attached to their motors. But those springs absorb and recycle energy that would be ordinarily lost with every step. The lightweight shins and thighs reduce shock loads when its legs swing and hit the ground. Like a high-end sports car, all of ATRIAS’ mechanisms are designed and tuned to cooperate with each other and enable maximum performance.
But also like a fast car, ATRIAS is hard to drive. The mathematics commonly used to control robotic walking just doesn’t work for ATRIAS. This means that researchers at the Dynamic Robotics Laboratory, in collaboration with Dr. Hartmut Geyer's laboratory at Carnegie Mellon, have been constantly inventing their own controllers to make ATRIAS go. Their goal: make ATRIAS walk, maneuver over obstacles, and run on the open outdoor stage.
With less than 100 days before ATRIAS’ big debut, the researchers are making fast progress. In fact, ATRIAS is busy tweeting its milestones and accomplishments every day on social media. To see all the latest videos and pictures, follow ATRIAS on Twitter @ATRIASrobot or on its Google+ and Youtube pages. Feel free to comment and tweet your questions... ATRIAS has been known to answer them personally.
ATRIAS is an acronym for Assume The Robot Is A Sphere. This represents our philosophy for designing ATRIAS, to make it as close to a simple spring-mass system as we can.
We want robots to be every bit as agile and robust at moving around as people. Bipedal robots still have a hard time reacting when they're tripped up with an obstacle they don't see coming. We work to make ATRIAS able to recover from curbs, pits, and puddles that it can't plan for in advance.
Bipedal robots, as a whole, are also very inefficient at walking. Studies have estimated that it costs humanoid robots more than 16 times as much energy to walk as it does a person. As such, bipedal robots can typically operate on battery for only about an hour. We know it's possible to be much more efficient at walking (people do it every day), and we believe making robots more dynamic is important to achieving this goal.
We have no plans to install motorized feet on ATRIAS, but it will have small passively-pivoting contact pads for traction. With ATRIAS, we were careful to include motors only where absolutely necessary. Motors have inertia, which means they can get in the way of the natural and dynamic motions that we designed ATRIAS to execute.
Carefully. For ATRIAS, it’s actually easier to hold position not by standing still, but by stepping in place. We call it dynamic standing.
It can and it has. You can see some videos of our collaborators in Michigan making great progress on that front. Dynamic 3D locomotion is tricky to test, so we develop controllers in 2D on the boom and extend them to 3D when we’ve worked out the bugs.
Update 2015-02-26: Take a look at some of our latest experiments in 3D!