摘要
本文从临机规划当前面临的难点问题出发,提出临机规划的智能发展需求,初步描述了临机规划问题,根据“快算”“盲评”“会变”三项能力要求,指出智能临机规划技术研究的关键特征是智能体设计与应用两个阶段的人机耦合,并明确了构建典型场景下可分层调度的任务框架、从军事专家到智能体的知识传导路径、兼容人工算法调用与机器自主学习的分层深度强化学习算法框架三点核心技术要求,基于当下智能博弈技术进展,提出了深度嵌入指挥员临机规划决策“OODA”回路各环节的智能临机规划技术设想。最后,从智能临机规划技术实现角度出发,提出了研究的总体目标、重点研究问题以及研究路线建议,并展望了未来智能临机规划技术的应用前景。
Grounded in the current problems in contingency planning,this paper puts forward the requirements for its intelligent development and preliminarily describes the issue of contingency planning.Based on the three capability requirements of rapid calculation,assessing with incomplete information,and appropriately adjusting to change,this paper highlights that the critical feature of intelligent contingency planning technology is human–machine coupling in both the design and deployment of intelligent agents.It then clarifies three core technical requirements—the task framework for constructing hierarchical scheduling in typical scenarios,the knowledge transfer path from military experts to intelligent agents,and a deep reinforcement learning algorithm framework for hierarchical task scheduling compatible with artificial model invocation and autonomous learning.Based on the current development of intelligent game technology,intelligent contingency planning technology that is deeply embedded in each link of the observe-orient-decide-act loop in commanders’contingency planning and decision-making is proposed.Finally,the paper analyzes the overall objective,key research issues,and recommended research routes from the perspective of the implementation of intelligent contingency planning technology,and sheds light on its application prospects in the future.
作者
包战
张驭龙
朱松岩
王春光
刘忠
BAO Zhan;ZHANG Yulong;ZHU Songyan;WANG Chunguang;LIU Zhong(Unit 31002,Beijing 100000,China;College of System Engineering,National University of Defense Technology,Changsha 410073,China)
出处
《国防科技》
2023年第1期112-118,共7页
National Defense Technology
关键词
作战规划
临机规划
人工智能
operational planning
contingency planning
artificial intelligence
