一种改进的非玩家角色行为树设计与实现
首发时间:2014-10-15
摘要:随着电脑游戏这种娱乐方式的逐渐发展,人工智能已经成为构建游戏交互内容的一种重要方式和技术,为了在游戏中构建具有足够说服力的人工智能角色,许多重要的技术被运用到游戏当中,近年来使用最频繁的行为树在大规模人工智能构建中优于层级状态机(HFSM)等早期使用的人工智能技术 ,但是行为树的构建需要具备丰富的业务经验和花费大量成本。本文提出一种基于增强学习算法(Q-Learning)的行为树实现方式,这种方法能够在行为树的构建过程中确认AI逻辑行为分支的最佳执行点,帮助分析与优化行为树的原型设计,实现自动化的行为树设计。
关键词: 人工智能 行为树 Q-Learning 电脑游戏 非玩家角色
For information in English, please click here
An enhanced behavior tree design and implementation for Non-Player Characters in computer games
Abstract:Artificial intelligence has become an increasingly important aspect of computer game technology to create interactional game content. In order to construct persuasive intelligent characters, more and more important technologies have been used in computer games. Recently behavior tree has been frequently used in this area, and it's more suitable than hierarchical state machine when scaled up, however, the design and creation of behavior trees requires experience and effort. This research introduces an enhanced behavior tree, which uses reinforcement learning as a method for tree design. This technology can assist game developers in identifying the most appropriate moment to execute each branch of AI logic, analyze and optimize early behavior tree prototypes and automate tree design.
Keywords: artificial intelligence behavior tree Q-Learning computer game Non-Player Character
基金:
论文图表:
引用
No.4611694100734614****
同行评议
共计0人参与
勘误表
一种改进的非玩家角色行为树设计与实现
评论
全部评论0/1000