Evolution of intraspecies communication in Q-Learning artificial organisms

摘要

An artificial life simulation is presented in which a population of artificial organisms evolutionarily synthesize an intraspecies communicationprotocol for mate-finding. These organisms cvolve by two kinds of evolution: Darwinian natural selection and Lamarckian inheritance of acquired traits. The more effectively an organism finds its mate, the more offsprings it produces. Not only recombining its successful parents' traits, but also an organism empirically improves its traits and passes thus acquired traits to its offspriings. By representing each of the mate-finding organisms as a Q-learning reactive agent augmented by this hybrid scheme of evolution, we show how steadily they can sysnthesize an appropriate communication protocol.