博弈智能的研究与应用

Game intelligence is a cross-disciplinary field that encompasses game theory and artificial intelligence. It focuses on interactions between individuals or organizations and the accurate solution of optimal strategies through quantifying game relationships with modeling, finally forming intelligent decision-making and its knowledge base. In recent years, with the explosion of massive behavioral data and the diversification of game forms, game intelligence has increasingly attracted the interest of researchers and has been used widely in real life. This paper systematically surveys game intelligence in three aspects. First, it reviews the relevant background of game intelligence, including single-agent Markov decision processes, multiagent modeling techniques based on game theory, and multiagent solution methods such as reinforcement learning and game learning. Second, based on the different game relationships between intelligent agents, this paper categorizes games into three paradigms: cooperative games, adversarial games, and mixed games, and introduces the main research problems, the mainstream research methods, and typical applications in each game intelligence paradigm. Finally, this paper summarizes the current research status of game intelligence, the main problems and research challenges that need to be addressed, and the prospects for application in academia and industry, providing a reference for related research and further promoting the development of the national artificial intelligence strategy.