Meeting Abstract

86.6  Thursday, Jan. 7  Simulating Evolutionary Processes: Swimming Robots in a Predator-Prey Ecology ROBERTS, S*; HIROKAWA, J; GUTIERREZ, A; ROSENBLUM, H; STICKLES, E; SAKHTAH, H; PORTER, M.E.; LIEW, C; ROOT, R; LONG, J; Vassar College; Vassar College; Vassar College; Vassar College; Vassar College; Vassar College; Vassar College; Lafayette College; Lafayette College; Vassar College soroberts@vassar.edu

To test hypotheses about how predation and foraging acted as selection pressures on early vertebrates, we created a predator-prey ecology in which a population of autonomous surface-swimming robots competed. Our aquatic ecosystem was populated by one predator robot, whose traits remained constant, and six prey robots, whose three evolvale traits were genetically coded: (1) span of the caudal fin, (2) number of vertebral elements, and (3) predator detection sensitivity. The prey robots were programmed to seek light, a behavior meant to imitate foraging. When a prey robot detected the prey-seeking predator, foraging behavior was over-ridden by an escape response. We evolved the population of six prey individuals over ten generations. Individual fitness was proportional to the distance from the predator, the inverse of the distance from the food source, peak acceleration during the escape, and the number of escape responses. The three individuals with the highest fitness were sexually reproduced in software using a genetic algorithm. We found significant directional increases in caudal fin span and number of vertebral elements, indicating that the selection pressures might have been sufficient to drive the evolution of these traits. Predator detection sensitivity showed a different pattern, with a significant directional decrease and oscillating variation, indicating that there may have been several predator detection sensitivities adaptive in our artificial ecosystem. This work was supported by NSF DBI-0442269 and IOS-0922605.