| 摘 要: 针对当前海空集群对抗智能指挥控制决策研究缺乏有效的算法研究平台的问题,设计了深度强化学习算法研究平台。首先分析了应用需求,其次对平台组成、硬件架构进行了设计描述,最后针对想定设计、算法训练和算法评价进行了子流程设计。平台设计采用仿真内核和显示模块分离、并行与分布式运行的模式,支持深度强化学习算法的快速训练和决策控制效果的验证评价,避免直接进行海空装备实物实验耗费大量时间和面临未知风险,有效提高了算法研发效率。 |
| 关键词: 集群对抗;深度强化学习;平台设计;并行与分布式运行 |
|
中图分类号: TP183
文献标识码: A
|
| 基金项目: 装备预先研究领域基金项目(61400010114) |
|
| Design of Deep Reinforcement Learning Algorithm Research Platform for Sea and Air Swarm Confrontation |
|
LIU Baohong
|
(Hunan Cangshu Aerospace Technology Co., Ltd., Changsha 410073, China)
liubh_vip@163.com
|
| Abstract: This paper proposes to design a deep reinforcement learning algorithm research platform to address the lack of effective algorithm research platforms for intelligent command and control decision-making in sea and air swarm confrontation. Firstly, the application requirements are analyzed and then the platform composition and hardware architecture are introduced. Finally, a sub-process is designed for scenario design, algorithm training, and algorithm evaluation. The platform design adopts a mode of separate simulation kernel and display module, and parallel and distributed operation, supporting rapid training of deep reinforcement learning algorithms and the verified evaluation of decision control effects. It avoids unknown risks and a large amount of time consumed by direct physical experiments of sea and air equipment, and it can effectively improve the efficiency of algorithm research and development. |
| Keywords: swarm confrontation; deep reinforcement learning; platform design; parallel and distributed execution |
