强化学习
计算机科学
空战
人工神经网络
状态空间
异步通信
人工智能
过程(计算)
代表(政治)
功能(生物学)
战场
动作(物理)
机器学习
模拟
计算机网络
历史
古代史
统计
物理
数学
量子力学
进化生物学
政治
政治学
法学
生物
操作系统
作者
Qiang Fu,Chengli Fan,Yong Heng
标识
DOI:10.1145/3580219.3580247
摘要
Aiming at the problems of the existing target allocation methods in practical application, such as insufficient representation of game antagonism and weak representation of tacit knowledge in the combat process, this paper studies the intelligent target assignment method based on deep reinforcement learning. Firstly, based on the operational characteristics of air defense target assignment, a new type of deep neural network for high-dimensional "state action" space is established, and the input and output information categories of the network, the state space and action space of each node are studied. The reward function is designed, and the strategy parameters are smoothly optimized by asynchronous training in the digital battlefield simulation environment by using the near end strategy optimization algorithm with tailoring. Simulation results show that the intelligent target assignment neural network model proposed in this paper has advantages and applicability.
科研通智能强力驱动
Strongly Powered by AbleSci AI