计算机工程与应用2026，Vol.62Issue(17)：18-34,17.DOI:10.3778/j.issn.1002-8331.2511-0177
智能博弈对手建模方法研究综述
Survey of Opponent Modeling Methods in Intelligent Gaming
摘要
Abstract
Opponent modeling is a key technique in intelligent gaming that improves decision-making in dynamic,uncer-tain environments by inferring and predicting adversary behaviors,strategies,and intentions.It has broad applications in domains such as game AI,military simulation,and cybersecurity.This paper systematically reviews fundamental concepts,major challenges,and methodological frameworks for opponent modeling.A classification scheme based on learning mechanisms organizes existing methods into three categories:parameter modeling,strategy modeling,and relation modeling.Each category is analyzed in terms of its core principles,representative algorithms,applicable contexts,and limitations.The paper also examines the trend toward hybrid modeling and identifies future research directions,including bounded rationality modeling,the integration of large language models,and considerations of interpretability and safety.This work provides a reference for ongoing research in the field.关键词
对手建模/智能博弈/深度学习/强化学习Key words
opponent modeling/intelligent gaming/deep learning/reinforcement learning分类
信息技术与安全科学引用本文复制引用
白金印,朱巍,徐庆林,叶晨浩,钟义豪,倪田晋..智能博弈对手建模方法研究综述[J].计算机工程与应用,2026,62(17):18-34,17.基金项目
国家社会科学基金(2023-SKJJ-B-107) （2023-SKJJ-B-107）
国防科技大学信息通信学院创新人才培育基金(YJKT-QT-25014). （YJKT-QT-25014）
