Aprendizaje de estrategias de decisión en juegos repetitivos no cooperativos

Abstract

This article presents the design and implementation of different mechanisms applied to evolutionary processes within non-cooperative strategies, especially applied to the iterated prisoner's dilemma (a widely-used reference model in the field of evolutionary economics). The strategies developed for the evolution mechanisms were Genetic Algorithms (GA), whereas Particle Swarm Optimization (PSO) was used for the evolution of game strategies. The result is a simulation environment that can be used to verify the emergence of strategies. Emergent strategies can defeat other strategies through a training process. In this en¬vironment games can be specified using a block programming approach or a textual domain specific language, facilitating the programming tasks involved to a great extent.

Keywords

• Genetic algorithms
• programming languages
• PSO
• game theory