AI Mimics Quadruped Gait Patterns

Researchers develop an artificial neural network that generates multiple distinct quadruped gaits and transitions between them.

Mar. 24, 2026 at 4:19am

Got story updates? Submit your updates here. ›

Researchers at Brown University's Carney Institute for Brain Science have developed an artificial neural network that can generate multiple distinct quadruped gait patterns, including bounding, pacing, trotting, walking, and pronking, as well as the ability to transition between these gaits. The network is based on attractor networks, a mathematical construct used to model neural activity patterns, and provides new insights into how the brain may process complex behaviors. The efficient and streamlined network could also be useful in advancing the technology of quadruped robots, enabling them to perform complex dynamic movements more autonomously.

Why it matters

This research expands the attractor network framework beyond static brain behaviors to include dynamic behaviors, providing a unified theoretical framework that can be used to study a range of brain functions. The findings could also inspire the development of more autonomous and efficient quadruped robots that can better mimic the complex gait patterns of four-legged animals.

The details

The artificial neural network developed by the researchers is based on a streamlined and efficient network of 24 artificial neurons. It is able to generate five distinct quadruped gaits - bounding, pacing, trotting, walking, and pronking - and capture the fast transitions between these gaits without needing to adjust any of the model's parameters. This suggests that attractor-based networks are more flexible and interpretable than other models for studying dynamic brain behaviors.

  • The research was published in the journal Neural Computation on March 24, 2026.

The players

Carney Institute for Brain Science

A research institute at Brown University that focuses on brain science.

Carina Curto

A professor of applied mathematics at Brown University and a co-author of the study.

Juliana Londono Alvarez

The lead author of the study and a postdoctoral researcher at Brown University.

Katherine Morrison

A professor of mathematical sciences at the University of Northern Colorado who collaborated with the research team.

Got photos? Submit your photos here. ›

What they’re saying

“We know the brain has to be able to flexibly and robustly maintain and change rhythms. By tapping into the rules of attractor networks, we have created an artificial neural network that hints at how biological brains might simultaneously encode and transition between different patterns and rhythms.”

— Carina Curto, Professor of Applied Mathematics

“This paper shows that you can expand attractor networks beyond the static to include the dynamic. Once you do that, you can see how the same principles underlying memory encoding can also generate something dynamic, like these gaits.”

— Juliana Londono Alvarez, Postdoctoral Researcher

What’s next

The researchers are currently in discussions with roboticists about adapting the neural network for use in quadruped robot projects, with the goal of enabling more autonomous and efficient dynamic movements.

The takeaway

This research demonstrates how attractor-based neural networks can be used to model not just static brain behaviors, but also dynamic and complex behaviors like quadruped gaits. The findings could lead to advancements in both our understanding of the brain and the development of more capable quadruped robots.