Abstract:

To address the timeliness requirements of large-scale weapon-target assignment， this paper proposed a solution model based on a multi-channel decision pointer network. First， a representation method based on weapon partition was designed to statically process dynamic feature dimensions， resolving the problem of varying feature dimensions. Second， a multi-channel decision model was constructed， utilizing context vectors and feature update mechanisms to enhance information interaction between channels and improve global solution capability. Simulation experiments demonstrate that this model significantly reduces solution time while maintaining accuracy， exhibits good generalization across problems of varying scales， and can effectively meet the timeliness and precision requirements for solving large-scale weapon-target assignment problems.

Key words: weapon-target assignment, combinatorial optimization, deep reinforcement learning, multi-channel decision, pointer network

摘要：

针对大规模武器目标分配的时效性需求，提出一种基于多通道决策指针网络的求解模型。通过设计基于武器划分的表征方式将动态特征维静态化处理，解决特征维度变化问题；构建了多通道决策模型，利用上下文向量和特征更新机制增强通道间信息交互，提升全局求解能力。仿真实验表明，该模型在保证精度的同时显著缩短求解时间，且对不同规模问题具有良好的泛化性，能够有效满足大规模武器目标分配问题求解的时效性与精度要求。

关键词: 武器-目标分配, 组合优化, 深度强化学习, 多通道决策, 指针网络