In this paper we present a novel agent-based simulation model for the natural selection of replicating agents whose survival depends on their programs for selecting their strategy to interact with each other. Game theoretic models describe this kind of interaction. The simulator can be used both for analysis, design and verification of autonomous systems (by intuitively abstracting them into our model and running the simulation). Good system design can be selected even for difficult, underspecified design problems. Although the inspiration of the model comes from evolutionary game theory, the evolving agents may represent not only biological, but also technical systems (e.g. software agents), and thus in some cases the underlying evolutionary mechanisms and observations differ from the typically published results.
展开▼