Organisms use sensory input when controlling their movement. Simultaneously, movement can enrich sensory input as the organism explores its environment. This talk will describe data driven analysis that analyzes motion capture data of a weakly electric fish engaged in a simple goal directed behavior. The results suggest that the fish switch between fast “explore” and slow “exploit” modes manifest by “broad shouldered” velocity distributions. Reanalysis of archived data from diverse species and behaviors finds that similar non-normal distributions are ubiquitous. Model simulations indicate that control algorithms mimicking these organismal behaviors outperform the prevailing standard of “persistent excitation” in adaptive control. Ongoing work will seek to further characterize the swimming behavior of these fish as stochastic, dynamical systems.

The seminar will meet in Rhodes Hall, Room 571.