基于深度学习的防空反导拦截决策研究

Air Defence and Anti-Missile Interception Decision-Making Study Based on Deep Learning

针对复杂海战场护航的任务场景,现有防空反导系统战术辅助决策功能在火控决策与武器火力分配方面分别具有对敌方模型依赖度高、拦截决策准确性差、无法有效利用战场历史数据和研究对象简单等问题,本文提出一种基于深度学习的反导拦截智能决策模型。首先,搭建战场仿真平台并分别对作战单元进行建模;然后,基于长短时记忆神经网络设计反导拦截智能决策模型;接着,利用匀速比例导引质点模型构建战前模拟数据以训练战前模型;最后,将战前模型迁移到战场模型中,并基于实际战场数据增强后的实时数据进行小样本在线训练。仿真结果表明,本文设计的反导拦截智能决策模型能够有效降低敌方模型依赖性,从而提升防空反导决策准确性。

In the scenario of complex naval escort missions,the current tactical decision support functions of antiaircraft missile defense systems face issues such as high dependency on enemy models,poor accuracy in interception decisions,inability to effectively utilize historical battlefield data,and simplistic research objects.To resolve the above problems,a deep learning-based anti-missile interception intelligent decision-making model was proposed in this study.Firstly,a battlefield simulation platform was established to model the combat units accordingly.Then,an anti-missile interception intelligent decision-making model was designed using Long Short Term Memory neural networks.After that,a pre-battle model was trained using simulated data acquired from a constant proportional guidance particle model.Finally,the pre-battle model was transferred to the battlefield model and fine-tuned with real-time data enhanced with actual battlefield data through small-sample online training.Experiment results show that the proposed anti-missile interception intelligent decision-making model can effectively reduce dependency on enemy models and improve the accuracy of air defense missile decision-making.