An Evolutionary Behavior Tool for Reactive Multi-agent Systems

Abstract

Multi-agent Systems (MAS) are a sub-area of Distributed Artificial Intelligence which focus on the study of autonomous agents and their actions in an environment. This paper presents a simulation environment for Reactive Multi-agent Systems called Simula++, where an evolutionary algorithm can modify the set of behavior rules of each agent. Our major goal is to define and develop a model to dynamically change the agents’ behavior in order to adapt the agents to their environment. In the Simula++ environment, an user can define a Reactive MAS where the predefined rules set of each agent can be modified to create new rules during simulation. That would happen through the precepts of Artificial Life and Evolutionary Algorithms.