horizontal line
Living Machines Home navigation buttonseparator lineOverview navigation barseparator linePeople navigation buttonseparator lineProjects navigation buttonseparator linePublications navigation bar

Robust Neural Modeling

People: Bryan Adams


The neurological control system for very simple animals can be a true exemplar of robust behavior. The flatworm notoplana acticola represents one of the most primitive brains, yet experiments have demonstrated that the brain can be removed, rotated 180 degrees, re-inserted, and functionality will be regained. Using this experiment as inspiration, this project seeks to use artificial evolutionary techniques to create homeostatic creatures in a variety of domains. Current preliminary research involves building a flexible neural evolution environment and developing a proper neural model. A vastly-simplified version of the flatworm's righting behavior is used as a testbed.


While many artificial neurons focus on finding a specific implementation that will yield a functional, evolvable controller, this project is concerned with looking for simple organizing principles that arise from evolved neural machines. By evolving controllers for a variety of tasks (using spike-train encoding), patterns between evolved controllers will point the way toward the design of robust control systems.

Homeostasis Powerpoint Presentation: powerpoint presentation link