Researchers in Carnegie Mellon University’s Robotics Institute (RI) have designed a system that makes an off-the-shelf quadruped robot nimble enough to walk a narrow balance beam.
“This experiment was huge,” said Assistant Professor Zachary Manchester. “I don’t think anyone has ever successfully done balance beam walking with a robot before.”
By leveraging hardware often used to control satellites in space, Manchester and his team offset existing constraints in the quadruped’s design to improve its balancing capabilities.
The standard elements of most modern quadruped robots include a torso and four legs that each end in a rounded foot, allowing the robot to traverse flat surfaces and even climb stairs. This new design resembles a four-legged animal, but, unlike animals, quadruped robots do not have instinctive agility. As long as three of the robot’s feet remain in contact with the ground, it can avoid tipping over. But if fewer than three feet are on the ground, the robot has a much higher risk of falling. Such a lack of balance makes walking over rough terrain particularly difficult.
“With current control methods, a quadruped robot’s body and legs are decoupled and don’t speak to one another to coordinate their movements,” said Manchester. “So how can we improve their balance?”
Their solution employs a reaction wheel actuator (RWA) — widely used in the aerospace industry to perform attitude control on satellites by manipulating the angular momentum of the spacecraft — system that mounts to the back of the quadruped robot. With the help of a novel control technique, the RWA allows the robot to balance independent of the positions of its feet.
“You basically have a big flywheel with a motor attached,” said Manchester. “If you spin the heavy flywheel one way, it makes the satellite spin the other way. Now take that and put it on the body of a quadruped robot.”
The team prototyped its approach by mounting two RWAs on a commercial Unitree A1 robot — one each on the pitch and roll axis — to provide control over the robot’s angular momentum. The RWAs provide independent control of the body’s orientation.
Manchester noted that it was easy to modify 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 limitations of a tail or spine. Without needing to account for such constraints, the hardware can be modeled like a gyrostat — an idealized model of a spacecraft — and integrated into a standard model-predictive control algorithm.
The team successfully demonstrated the robot’s enhanced ability to recover from sudden impacts. In simulation, they mimicked the falling-cat problem by dropping the robot upside down from nearly half a meter, with the RWAs enabling the robot to reorient itself mid-air and land on its feet. On hardware, they showed the robot’s ability to recover from disturbances — as well as the system’s balancing capability — with an experiment where the robot walked along a six-centimeter-wide balance beam.
With continued work to enhance a quadruped robot’s stabilizing capabilities to match the instinctual four-legged animals that inspired their design, they could be used in high-stakes scenarios like search-and-rescue in the future.
“Quadrupeds are the next big thing in robots,” said Manchester. “I think you’re going to see a lot more of them in the wild in the next few years.”
For more information, contact Aaron Aupperlee at
