Nimble and steady on their feet: Next-generation legged robots

by

Drake, an open-source model-based
design and verification robotics
simulator, helps verify Kurtz's models.

Drake, an open-source model-based design and verification robotics simulator, helps verify Kurtz's models.

Could legged robots be used to find survivors after an earthquake or to fight forest fires? Answering that question is the goal of Vince Kurtz’s research.

A graduate student in the Department of Electrical Engineering, Kurtz recently received a three-year Dolores Zohrab Liebmann Fellowship supporting his efforts to bridge the gap between classical control theory, formal methods and robotics.

“Today’s best legged robots work well under certain conditions, but they perform poorly in unstructured environments outside the lab,” Kurtz said.

“If we want to be able to use them in dangerous and uncertain environments, robots must be able to better mimic the way humans and animals walk, especially over uneven ground when balance is critical.”

Kurtzvince2
Vince Kurtz

Current control algorithms — mathematical instructions written into code to achieve a task, like avoiding obstacles or balancing without falling — for legged locomotion are complex and require extensive hand-tuning by the robot’s operators.

Kurtz connects the simple models most roboticists use to control legged robots with more complete physics-based models. These mathematical connections enable robots to recover from larger push disturbances and walk over more difficult terrain. The connections may even provide clues to how animals move so effectively over land.

Kurtz is continuing his work with his adviser, Hai Lin, professor of electrical engineering and Patrick Wensing, assistant professor of aerospace and mechanical engineering, to design safe and effective control methods for humanoid and quadruped robots.