基于深度强化学习的体系设计空间探索方法  

System of Systems Design Space Exploration Based on Deep Reinforcement Learning

作　　者：陈涛[1] 林萌龙 陈洪辉[1] 张萌萌[1] CHEN Tao;LIN Menglong;CHEN Honghui;ZHANG Mengmeng(Science and Technology on Information Systems Engineering Laboratory,National University of Defense Technology,Changsha 410073,China)

机构地区：[1]国防科技大学信息系统工程重点实验室,长沙410073

出　　处：《火力与指挥控制》2022年第12期109-114,共6页Fire Control & Command Control

摘　　要：当前的作战样式正逐步从以网络为中心向以决策为中心转变,以“马赛克战”为代表的分布式作战理念极大增加了体系设计的复杂程度,同时也为作战体系设计在实时性和动态性等方面提出了新要求。人工智能技术为实现作战体系架构的智能辅助设计提供了新手段。从当前体系设计面临的问题出发,提出了基于深度强化学习的体系设计空间探索模型,介绍了其中涉及的关键技术,并以体系通信网络结构的设计为例展示了所提方法的有效性。The current combat mode is gradually shifting from network-centric warfare to decisioncentric warfare.The distributed combat concept represented by“mosaic warfare”increases the complex degree of SoS design greatly and puts forward new requirements for combat SoS design in terms of realtime and dynamics,etc.Artificial intelligence technology provides a new means to realize the intelligent auxiliary design of the combat SoS architecture.Starting from the problems faced by the current SoS design,a SoS design space exploration model based on deep reinforcement learning is proposed,the key technologies involved are introduced,and the design of the SoS communication network structure is taken as an example to demonstrate the effectiveness of the proposed method.

关 键 词：体系架构 深度学习 强化学习 设计空间 

分 类 号：TN929[电子电信—通信与信息系统]