A Soft Computing Framework for Adaptive Agents

We extend the actor-based model of concurrent and distributed programming toward a framework in which the agents can learn emerging behaviours through a fuzzy evolutionary structure. The framework is based on the notion of FuzzyEvoAgent, i.e. an entity that exploiting the basic issues of actors (asynchronous message passing, concurrent computation) is submitted to evolutionary laws that reinforce the most suitable behaviors respecting the environment. The behavior evolution is accomplished without the intervent of external stimulus, so that the actor (an entity that reacts only if it receives a command) may be considered as an adaptive agent. We propose a formal definition as well as an implementation model of FuzzyEvoAgents that has been verified via a simple simulation of Artificial Life.