A First Order Language to Coevolve Agents in Complex Social Simulations



Principles from the sciences of complexity may be applied to the problem of generating interesting and surprising high-level behaviours in 3D world simulations. We present a first order language designed to represent agents' internal reasoning rules that is suitable for a coevolutionary environment. An algorithm is described to use a set of rules expressed in this language to produce decisions. Genetic operators are defined. The first experimental results are presented, showing that agents implementing our model are able to develop interesting behaviours under open-ended evolution. In this initial results, the agents show the ability to bootstrap their own evolutionary process, by developing reproductive technology.


Evolutionary and Complex Systems


European Conference on Complex Systems 2006, September 2006

Cited by

No citations found