Developing Human-Like Game Agents with a Wide Range of Play Styles and Skill Levels

近年来出现了在日益复杂的游戏中表现出超人技能的AI代理，但这些代理并不总是对游戏开发者有用。AI代理可以代表玩家的技能水平和游戏风格，从而在游戏开发过程中提供帮助。拟议的研究将集中在三个领域：测量技能和游戏风格，开发模仿一系列人类游戏风格和技能水平的游戏代理，以及使这些代理像人类一样。在成功完成后，这项研究有可能通过自动化游戏测试来改进开发过程，并引入更具吸引力和互动性的AI代理。

AI agents that perform with superhuman skill in increasingly complex games have appeared in recent years, but these agents are not always useful to game developers. AI agents can aid in the game development process by representing the skill levels and play styles of players. The proposed research will focus on three areas: measuring skill and play styles, developing game agents that mimic a range of human play styles and skill levels, and making these agents human-like. Upon successful completion, this research has potential to improve the development process via automated playtesting and to introduce AI agents that are more engaging and interactive.