基于改进UCT算法的爱恩斯坦棋博弈设计研究  

Research on the Design of Einstein̓s Chess Game Based on Improved UCT Algorithm

作　　者：孙嘉明 胡光元[1] 王佳宁 岳海龙[1] 张杰斯 SUN Jiaming;HU Guangyuan;WANG Jianing;YUE Hailong;ZHANG Jiesi(Shenyang Aerospace University,Shenyang 110136,China)

机构地区：[1]沈阳航空航天大学,沈阳110136

出　　处：《移动信息》2024年第10期246-248,共3页Mobile Information

摘　　要：为了解决传统UCT算法在爱恩斯坦棋中评估精度低和搜索效率低的问题,文中以基于UCT算法的爱恩斯坦棋为研究对象,通过基于修正值的收益值计算方法和并行优化处理,对传统UCT算法的估值方法和线程问题进行了改进,实现了在有限时间内达到更优的搜索精度和搜索效率。结果表明,改进的多线程UCT算法在修正系数为2.5时的收益值以及胜率明显大于传统单线程UCT算法,且在搜索效率上具有显著提高。In order to solve the problems of low evaluation accuracy and low search efficiency of traditional UCT algorithm in Einstein chess,this paper takes Einstein chess based on UCT algorithm as the research object,and improves the valuation method and threading problem of traditional UCT algorithm through a profit value calculation method based on correction value and parallel optimization processing,achieving better search accuracy and efficiency in a limited time.The results show that the improved multi-threaded UCT algorithm has significantly higher profit and win rates than the traditional single threaded UCT algorithm when the correction coefficient is 2.5,and has a significant improvement in search efficiency.

关 键 词：计算机博弈 人工智能 传统UCT算法 改进UCT算法 爱恩斯坦棋 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]