The State-Action-Reward-State-Action Algorithm in Spatial Prisoner's Dilemma Game

• DOI: 10.48550/arxiv.2406.17326
• Publication Date: 2024-06-25
• Authors: Lanyu Yang; Dongchun Jiang; Fuqiang Guo; Mingjian Fu
• Keywords: Superrationality

Cooperative behavior is prevalent in both human society and nature. Understanding the emergence maintenance of cooperation among self-interested individuals remains a significant challenge evolutionary biology social sciences. Reinforcement learning (RL) provides suitable framework for studying game theory as it can adapt to environmental changes maximize expected benefits. In this study, we employ State-Action-Reward-State-Action (SARSA) algorithm decision-making mechanism theory. Initially, apply SARSA imitation learning, where agents select neighbors imitate based on rewards. This approach allows us observe behavioral without independent abilities. Subsequently, utilized primary independently choose or betrayal with their neighbors. We evaluate impact rates by analyzing variations rewards distribution cooperators defectors within network.