유전 알고리즘을 적용한 확장된 죄수의 딜레마 게임에서 선택확률함수 조정에 따른 유효한 전략 분석

Abstract

The Prisoner’s Dilemma, a representative example of game theory, is a classic game in which two players receive rewards based on two action strategies(cooperation or defection). However, many variations have been made depending onchanges in conditions(number of participants, behavioral strategies, reward values, number of games, etc.). In this study, an extended Prisoner’s Dilemma game was used in which a genetic algorithm was applied to the iterative Prisoner’sDilemma game to improve the selection probability function to determine which action strategy should be selected according to changes in reward value. A possible SPS(selection probability strategy) was developed. As a result of theprogram simulation, this SPS ranked high in the early stages, showed good performance, and was generally a dominant strategy that survived until the final stage. This study is significant in that it suggests the possibility of developing a new strategy in the Prisoner’s Dilemma game and reaffirms existing strategies that are generally effective.