Two-level evolution of foraging agent communities

Abstract

This paper presents simulation results of artificial foraging agent communities. The goal of each agent in the community is to find food. Once a food source is found, agents eat portions of it and carry some other portions to the nest (in a manner similar to ants) until the food is depleted. Agents may also communicate food positions when they are near each other. They are given a set of genes that control several characteristics, such as their activity, memory, scepticism, lying, etc. These genes are recombined and propagated by means of sexual reproduction. When one nest is superpopulated with agents, it can break in two nests. Agents can communicate only with those belonging to the same nest, which gives rise to emergent situations of competition and cooperation between the agents in the same nest, as well as competition between different nests. Other emergent phenomena such as the propagation of rumours are also studied.