Evolutionary game theory combined with reinforcement learning synthesis - A comprehensive survey

Within the realm of modern intelligent science, the fields of evolutionary game theory (EGT) and reinforcement learning (RL) have exhibited an inherently intertwined relationship since their inception. This symbiotic association has been further accentuated, especially in light of the significant advances in artificial intelligence technologies witnessed in recent years. As a result, their synergistic interactions have grown increasingly profound, catalyzing an explosive surge in interdisciplinary research endeavors. To align with this prevailing trend, the objective of this survey is to provide a comprehensive assessment of the interconnections, distinctions, and recent advances in the combination of EGT with RL. We approach this from three perspectives: Q-learning, the Bush-Mosteller (BM) model, and deep reinforcement learning (DRL). Subsequently, the most recent advances in research within these three dimensions are categorized based on distinct subject areas. Ultimately, this nascent cross-disciplinary research domain is comprehensively summarized, and its future prospects are outlined.