Research Videos

Subscribe to be notified of future videos

Overview of the EvolvingAI lab.

Evolving neural networks (digital brains) to control how robots walk.

Why do modules evolve in biological networks? In this video we show that the presence of a cost for network connections leads to the evolution of modules and increases performance. This discovery helps answer a biological mystery and provides a boost to artificial intelligence research, because it provides a technique for evolving modular artificial neural networks. The networks with a connection cost are more modular and have higher fitness. Click here for more information.

Evolving soft robots with multiple-materials and a generative encoding based on developmental biology.

Encouraging Creative Thinking in Robots: The Creative Thinking Approach

Evolving Neural Networks That Are Both Modular and Regular: HyperNeat Plus the Connection Cost Technique. Design objects with evolution and 3D print them! Nearly 4 million objects have been evaluated to date.

A summary of my recent research into evolving neural networks that have the properties of regularity (e.g. symmetries and repeated structures) and modularity (e.g. neural modules).

Evolved Electrophysiological Robots

Novelty Search Creates Robots with General Skills for Exploration

Harlem shake: evolutionary robotics edition. (Done by one of my students)

My dissertation presentation on evolving neural networks with generative encodings based on developmental biology.

Aracna: An open-source, 3D printed robot for testing artificial intelligence algorithms.

Watch the Aracna robot be 3D printed.