Um protótipo de software para simulação de sistemas multiagentes baseados na abordagem de Russell/Norvig e na teoria dos jogos

Abstract

The basis for the development of artificial intelligence is based on a primordial concept - intelligent agents. Studying, understanding and improving the behavior of these autonomous entities are some of the motivations of this work. Understanding and improving the behavior of these autonomous entities are some of the motivations of this work, which seeks to present a software prototype, which enables the study of agent behaviors, using the structure proposed by Russell and Norvig and introducing a way of classifying the strategies of behaviors used in Prisoner's Dilemma games. It is used as a basis for Axelrod's computational tournaments in 1984, as well as other tournaments that took place in these more than 30 years of research into the Iterated Prisoner's Dilemma. As an integral part of this work, we used Vincent Knight's library, which features 242 behavior strategies implemented in Python. In addition to the library, 512 more strategies were added, thus forming a base of 754 strategies all implemented in Python. By uniting the field of artificial intelligence with game theory, the aim is to use the best of both areas for application in highly complex problem solving.