CMU taught a robot dog to walk a balance beam

by Ana Lopez

When I was discussing humanoid robots not too long ago, someone told me that their main problem with the form factor is that – from an evolutionary point of view – we are not particularly well built. Of course, that’s not to say that our bodies haven’t served us well.

They’ve been doing the trick for a few hundred thousand years. It’s more that if you sat down with a talented product designer and asked them to come up with something from scratch, certain concerns would probably lead them in an entirely different direction.

Balance is on that list. Again, we’ve done just fine for ourselves, all things considered, but if balancing was high on your priority list, you could opt for something with four legs and a lower center of gravity.

This ready-made dog robot is a great place to start. The quadruped is sufficiently stable in its standard locomotion. As you’d probably expect, that changes quickly if you, for example stick it on a balance beam. That’s the kind of challenge you live for, though, when you’re part of a lab like Carnegie Mellon University’s Robotics Institute.

“This experiment was huge,” says assistant professor Zachary Manchester. “I don’t think anyone has ever successfully walked a balance beam with a robot before.”

Why this is such a big challenge… these robots aren’t designed for it to begin with. Again, if you’re that almighty designer, you’d add more flexibility and counterbalance to begin with. The solution the team landed on is that large backpack you see in the photo above. That’s a reaction wheel actuator (RWA) – something used to help control the altitude of satellites.

“You basically have a big flywheel with a motor attached to it,” adds Manchester. “If you rotate the heavy flywheel in one direction, the satellite rotates in the other direction. Now take that and place it on the body of a four-legged robot.

CMU Notes:

Manchester said it was easy to adapt an existing control framework to account for the RWAs because the hardware doesn’t change the robot’s mass distribution, nor does it have the joint constraints of a tail or spine. Without having to deal with such limitations, the hardware can be modeled as a gyrostat (an idealized model of a spacecraft) and integrated into a standard model-predictive control algorithm.

Why, you may ask, would anyone spend time developing something like this? Aside from the obvious satisfaction of watching a dog robot walk a balance beam, the most immediate answer is search and rescue. That has long been an important application for these kinds of robots: to send machines where you would not normally send people. It’s easy enough to see why balance is super important in such a scenario.

Related Posts