摘要
在人工智能飞速发展的大背景下,游戏AI的革新亦走在技术的前沿并引领游戏产业发展。为了研究游戏AI产生更为高效的游戏策略,对游戏研发中创新上限的瓶颈进行突破,本文设计一款ML-Agents驱动的足球竞技游戏,通过游戏AI的学习过程增强游戏NPC的可适应性和拟人化。借助CINEMA 4D、Maya设计场景建模,使用Unity引擎构建游戏场景,运用Anaconda搭建虚拟训练环境,采用TensorFlow训练NPC的行为,应用ML-Agents插件构建程序化人格,训练并引导Agent产生所期望的行为。实践结果表明,采用该方案研发的足球竞技游戏案例产品,游戏AI产生了更具创新性、更具效能的行为模式。
Under the background of the rapid development of artificial intelligence,the innovation of game AI is still at the forefront of technology and leads the development of game industry.In order to study the game AI to produce more efficient game strategies,and break through the bottleneck of innovation ceiling in game research and development,this project designs a competitive football game driven by ML-Agents to enhance the adaptability and personification of game NPC through the learning process of game AI.The project uses Cinema 4D and Maya to design scene modeling,uses Unity engine to build game scene,and applies Anaconda to construct virtual training environment.Besides,it uses TensorFlow to train the NPC behavior,uses ML-Agents plug-in to build programmatic personality,as well as trains and guides agent to produce desired behavior.The practical results show that the football competitive game case product developed by this scheme makes the game AI produce a more innovative and efficient behavior mode.
作者
李昶
顾汉杰
LI Chang;GU Hanjie(College of Information Technology,Zhejiang Shuren University,Hangzhou,China,310015)
出处
《福建电脑》
2022年第1期81-84,共4页
Journal of Fujian Computer
基金
国家大学生创新创业训练计划(No.202011842027)资助。
