基于变分推断的高超声速滑翔目标跟踪方法  

Hypersonic glide target tracking method based on variational inference

作　　者：魏振伟[1,2] 梁彦 徐林峰[2] 詹文超 吉瑞萍 Wei Zhenwei;Liang Yan;Xu Linfeng;Zhan Wenchao;Ji Ruiping(Zhengzhou University of Aeronautics School of Automation,Zhengzhou 450015,China;School of Automation,Northwestern Polytechnical University,Xi'an 710129,China)

机构地区：[1]郑州航空工业管理学院自动化学院,郑州450015 [2]西北工业大学自动化学院,西安710129

出　　处：《战术导弹技术》2025年第1期104-112,共9页Tactical Missile Technology

基　　金：国家自然科学基金(61873205);河南省科技攻关项目(192102210103)。

摘　　要：针对非合作高超声速滑翔目标跟踪中动力学模型参数未知且时变的问题,提出了一种基于变分推断的自适应跟踪方法。先在半速度坐标系中建立目标的空气动力学模型,将机动控制量增广到状态中,把模型中的未知参数视为随机变量;然后,利用变分推断自适应滤波策略对目标运动状态和模型中的未知参数的概率密度函数进行联合估计和辨识。两种典型机动轨迹的跟踪仿真结果表明,所提方法能够较好地适应阶跃机动和连续幅值机动,跟踪精度优于基于运动学的自适应跟踪方法。An adaptive tracking method based on variational inference for non-cooperative hypersonic glide targets is proposed to deal with the unknown and time-varying model parameters. The aerodynamic model of hypersonic glide target is established in semi-speed coordinate system. The two control variables attack angle and bank are extended to the state component, and the unknown parameters in the model are treated as random variables. A variational inferential adaptive filtering method is adopted to jointly estimate and identify the motion states and unknown model parameters. The tracking simulation results of two typical maneuver trajectories indicate that the proposed method is well suitable for the step maneuver and the continuous maneuver, and the tracking accuracy is better than that of the kinematic-based adaptive tracking method.

关 键 词：高超声速滑翔目标 空气动力学模型 自适应滤波 参数估计 变分推断 KL散度 密度估计 

分 类 号：TN957[电子电信—信号与信息处理]