联合战役仿真环境对强化学习的挑战

Challenges to Reinforcement Learning in Joint Operational Simulation Environment

随着现代战争形态的不断演化,复杂战场环境对作战指挥提出了新的挑战。快速有效地决策成为未来战争中制胜的关键,应对之策只能依靠智能方法。传统基于事先设计的决策原则或初级人工智能辅助的方法均不能满足要求。近年来游戏智能的发展初步表明,深度强化学习技术在复杂环境下的决策领域中显现出巨大的潜力。但将其直接应用于作战决策问题研究时首先面临环境适配问题,即战役仿真环境和已被实践证明强化学习技术可以发挥作用的环境之间存在巨大鸿沟。从作战决策问题研究本身和联合战役仿真特点出发,剖析将通用的深度强化学习技术应用于联合战役仿真这一特殊环境需要面临的困难和挑战,为基于强化学习研究作战决策提供参考。

With the continuous evolution of modern warfare,the complex battlefield environment poses new challenges to the operational command.Quick and effective decision-making becomes the key to victory in future wars,and countermeasures can only rely on intelligent methods.Traditional decision-making principles based on pre-designed or primary artificial intelligence-assisted methods cannot meet the requirements.In recent years,the development of game intelligence has initially shown that deep reinforcement learning technology has shown great potential in the field of decision-making in complex environments.However,when it is directly applied to the research of operational decision-making problems,it faces the problem of environmental adaptation,i.e.,there is a huge gap between the battle simulation environment and the environment in which reinforcement learning technology has been proven to be effective.Starting from the research of combat decision-making problems and the characteristics of joint battle simulation,this paper analyzes the difficulties and challenges that applying general deep reinforcement learning technology to the special environment of joint battle simulation,and provides a reference for researching operational decision-making based on reinforcement learning.