Emergence of Different Mating Strategies in Artificial Embodied Evolution

  • Authors:
  • Stefan Elfwing;Eiji Uchibe;Kenji Doya

  • Affiliations:
  • Neural Computation Unit, Okinawa Institute of Science and Technology, Okinawa, Japan 904-2234;Neural Computation Unit, Okinawa Institute of Science and Technology, Okinawa, Japan 904-2234;Neural Computation Unit, Okinawa Institute of Science and Technology, Okinawa, Japan 904-2234

  • Venue:
  • ICONIP '09 Proceedings of the 16th International Conference on Neural Information Processing: Part II
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

In nature, discrete alternative mating strategies have been observed in many different species. In this paper, we investigate the emergence of different mating strategies in a small colony of simulated robots, using our previously proposed framework for performing embodied evolution with a limited number of robots. The virtual agents can reproduce offspring by mating, i.e., an exchange of genotypes with another robot. In the experiments, we observed two individual mating strategies: 1) Roamer strategy, where the agents never wait for potential mating partners; and 2) Stayer strategy, where the agents wait for potential mating partners depending on their internal energy level, the distance to the mating partner's tail-LED, and the distance to the closest battery. The most interesting finding was that in some simulations the evolution produced a mixture of mating strategies within the population, typically with a narrow roamer subpopulation and a broader stayer subpopulation with distinct differences in genotype, phenotype, behavior, and performance between the subpopulations.