基于改进TICC聚类算法的智能博弈宏观策略生成技术  

Macro Strategy Generation Technology of Intelligent Game Based on Improved TICC Clustering Algorithm

作　　者：周盼 程健庆[1] 初阳[1] ZHOU Pan;CHENG Jianqing;CHU Yang(Jiangsu Automation Research Institute,Lianyungang 222006)

机构地区：[1]江苏自动化研究所,连云港222006

出　　处：《舰船电子工程》2023年第2期69-74,共6页Ship Electronic Engineering

摘　　要：随着军事智能技术的发展,采用智能博弈技术构建决策智能体已成为热点研究方向。当前,以深度强化学习算法为主构建的决策智能体输出主要是基于实时态势的微观决策指令,而作战决策人员需要的往往是未来一段时间内的宏观策略。针对智能博弈中微观决策指令周期短,战场态势数据规模大、维度高,导致难以对智能博弈中宏观策略进行提取的问题,提出了一种基于时间序列的改进TICC分割聚类算法,对智能博弈形成的数据进行分割聚类,生成宏观策略,为作战方案的制作提供支撑。With the development of military intelligence technology,using intelligent game technology to build decision agent has become a hot research direction.At present,the output of decision agent mainly based on deep reinforcement learning algorithm is mainly micro decision instructions based on real-time situation,while the operational decision makers often need macro strategies in a period of time in the future.In order to solve the problem that it is difficult to extract the macro strategies in intelligent games due to the short cycle of micro decision-making instructions,large scale and high dimension of battlefield situation data in intelligent games,an improved TICC segmentation clustering algorithm based on time series is proposed to segment and cluster the data formed in intelligent games,generate macro strategies,and provide support for the production of battle plans.

关 键 词：聚类分析 智能博弈 宏观策略 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]