针对传统意图预测模型无法满足体系对抗下防空作战对预测准确性和可靠性的要求,提出基于 BiLSTM-Attention(bi-directional long short-term memory-attention)和动态贝叶斯网络的防空目标智能意图预测方法.综合考虑目标状态信息和目标战术信息,设计防空目标作战意图特征集;充分利用收集的历史时刻数据和预测的未来时刻数据,引入双向循环和注意力机制,模拟决策人员对于作战情况的推理过程,突出影响轨迹类型的关键信息,以提高防空目标轨迹预测的准确率.在此基础上,结合目标轨迹、类型、高度和速度,构建动态贝叶斯网络,以实现对防空目标意图的精准预测.仿真实验考虑了雷达开关机对目标轨迹的影响,通过与传统目标意图预测方法对比,验证了博弈对抗条件下所提方法的可行性和优越性.
Intelligent intent prediction of air defense targets based on BiLSTM-attention and dynamic Bayesian networks
Aiming at the fact that the traditional battle intent prediction model cannot meet the requirements of prediction accuracy and reliability for air defense operations under sys-tem confrontation,an intelligent battle intent prediction method for air defense targets based on BiLSTM-Attention(bi-directional long short-term memory-attention)and dynamic Bayesian network is proposed.Considering target state information and target tactical information com-prehensively,the combat intent feature set of the air defense target is designed;making full use of the collected historical moment data and predicted future moment data,the two-way loop and attention mechanism is introduced to simulate the reasoning process of the decision maker on the combat posture,highlighting the key information that affects the type of ballistic trajectory,and increasing the accuracy of the ballistic trajectory prediction of the air defense target.On this basis,a dynamic Bayesian network is constructed by combining the target trajectory,type,altitude and velocity to realize the accurate prediction of air defense target intent.The simula-tion experiments consider the effect of radar switch on target trajectory,and verify the feasibility and superiority of the proposed method under game confrontation conditions by comparing it with the traditional target intent prediction method.