Industrial Robots Could Be Programmed with Motion-Capture Systems

The distinction between collaborative and conventional industrial robots is their ability (or lack thereof) to work side-by-side with humans.

Part of the challenge in developing a robot that can work near its biological counterparts is establishing a system in which the robot can perceive its environment.

Cobots (short for collaborative robots) like Baxter from Rethink Robotics and the UR10 robotic arm from Universal Robots perceive their environment through proximity sensors. Cobots and their conventional predecessors are taught how to perform tasks through a controller or by being directly moved through the motions.

But what if a robot could actually see and a learn a process visually?

Madeline Gannon, a Ph.D. candidate at Carnegie Mellon University’s School of Architecture, designed a system for just that.

Quipt is an open-source software that allows a robot to perceive human motions as instructions using a motion-capture system. Optical sensors track markers made from retroreflective tape on a person’s hand or clothes to identify what target to focus on for instruction.

The robot can be ordered to follow, mirror or avoid markers. This allows engineers to have an intuitive understand of how the robot will move along with them.

Gannon developed the motion-capture system after working with industrial robots at CMU for a number of years.

"My research is really playing in the field of computer science and robotics, but the questions I'm able to ask in those specific domains is conditioned by my architectural background,” said Gannon. “It's really a spatial answer, how to control or interact with a robot. That, in my mind, is an architectural answer to this problem."

Golan Levin, director of the Frank-Ratchye STUDIO for Creative Inquiry at CMU, is one of Gannon's doctoral thesis advisors. He thinks her work could change how people design architecture, clothing and furniture, as well as influence industrial design and the arts. 

"Madeline is remarkable for the way in which she brings together an acutely sensitive design intuition with a muscular ability to develop high-performance software," Levin said. "The kind of work she is doing could not be achieved by a collaboration between a designer and engineer; it takes a single person with a unified understanding of both."

For more information on Gannon’s Quipt project, visit madlab.cc/quipt/.