Meeting Abstract

P3.109  Wednesday, Jan. 6  Evolving Intelligence in Autonomous, Fish-like Biorobots: Does Competition for Resources Matter? DE LEEUW, J.R.*; PORTER, M.E.; LIVINGSTON, K.R.; LONG, J.H.; Vassar College; Vassar College; Vassar College; Vassar College jodeleeuw@vassar.edu

We are interested in understanding how different selection environments drive the evolution of intelligence in aquatic animals. Because many of the character changes in early vertebrates appear to have enhanced mobility, we contrasted selection environments in which resources were scarce or bountiful. If greater intelligence, relative to other individuals in the population, is required to forage in a scarce-resource environment, then the population will evolve greater foraging intelligence compared to the population in the bountiful-resource environment. To model this system, we evolved two populations, one in each resource environment. Each population consisted of ten autonomous and physically-embodied biorobots, called NERDs (NeuroEvolving Robotic Device). Individuals differed only in their genomes, which coded for the topology of a neural network. The neural network converted sensor inputs from an infrared proximity detector array and from a photocell array into motor output at the flapping tail. The fitness function was the difference between the amount of photic energy harvested and the energy used to swim. Preliminary results indicate that a variety of foraging strategies evolve. This work was supported by NSF DBI-0442269 and IOS-0922605.