Meeting Abstract

P3.47  Tuesday, Jan. 6  Character Evolution in Robotic Fish ROSENBLUM, H.G.*; GUTIERREZ, A; ROBERTS, S; HIROKAWA, J; PORTER, M; LONG, J.H.; Vassar College; Vassar College; Vassar College; Vassar College; Vassar College; Vassar College harosenblum@vassar.edu

We use robots to test adaptation hypotheses about the evolution of three vertebrate characters: caudal fin span, number of vertebrae, and sensitivity of the lateral line. Is it plausible that the evolution of these characters was driven by selection for enhanced foraging and predator avoidance in a simple predator-prey ecology? To address this question, we evolved ten generations of a population of biologically-inspired, aquatic, and autonomous robots. In each generation, each of six individual prey robots swam in three trials, attempting to forage and, at the same time, avoid a pursuing predator robot. Relative individual fitness was measured using a compound function that rewarded faster cruising speed, greater escape acceleration, shorter distance from the food source, greater distance from the predator, and larger number of escape responses. Based on relative fitness, the top three individuals were allowed to mate, with a genetic algorithm mutating the genes, creating gametes, and making the new genotypes of the offspring. We hypothesized that all of our characters would increase in magnitude, evolving directionally and in concert, under constant selection. While caudal fin span and number of vertebrae evolved in a positive direction, as predicted, sensitivity of the lateral line appeared to undergo first stabilizing and then disruptive selection. Thus it appears plausible that separate characters can evolve both in concert and independently under a single selection regime. This work was supported by NSF DBI-0442269.