Gracias, Nuno, Henrique Pereira, José Allen Lima, and Agostinho Rosa. “Gaia: An artificial life environment for ecological systems simulation.” In ALIFE V , pp. 124-131. 1997.
URL1 URL2
This paper presents an ecology simulator for the study of certain aspects of ecology and biology, such as learning, evolution and population dynamics. The simulator is an artificial world, where two kinds of species can evolve: autotrophs and heterotrophs. Heterotrophic individuals, or critters, are capable of moving, eating, fighting and mating. They have a simple nervous system, a neural network, with a retina input. Associative Hebbian learning is used in the modification of synapses. Nervous system structure and physiological characteristics are coded in the critter’s genome. Autotrophs are static. They are born and grow according to a definable geographic distribution and rate. Simulations were carried out to study learning and behaviour evolution, in an approach as close as possible to biological reality. It was found that critter learning was essentially phylogenetic, i.e., Hebbian learning was limited to develop genetically defined connections, and did not perform any significant correlation between inputs and outputs. This suggests that, for animals with very simple nervous systems, behaviours evolve mainly through genetics, as opposed to a possible emergent scheme of reinforcement learning. Interspecific resource competition was also studied. It was found that if two similar species were sowed in the world, one of them won the competition. Several explanations for this phenomena are presented. A strong relation was found, between the geographic distribution of autotrophs and heterotrophs. Limit cycles of the two populations were observed.