基于Transformer架构的高超声速飞行器轨迹生成与预测算法  

An Algorithm for Trajectory Generation and Prediction of Hypersonic Vehicle Based on Transformer Architecture

作　　者：狄子琦 王翔宇 吴双 周宇[1] DI Ziqi;WANG Xiangyu;WU Shuang;ZHOU Yu(School of Electronic Engineering,Xidian University,Xi’an 710071,Shaanxi,China)

机构地区：[1]西安电子科技大学电子工程学院,陕西西安710071

出　　处：《空天防御》2023年第4期35-41,共7页Air & Space Defense

基　　金：国家自然科学基金(62376205)。

摘　　要：随着高超声速飞行器打击水平的不断增强,如何对其进行精确轨迹预测成为防守方的研究焦点。本文结合Transformer深度学习模型,提出基于参数估计的轨迹预测模型。首先,从空气动力学视角建构影响飞行器机动的控制参数模型,并归纳不同机动模式下控制参数的变化规律;其次,建立基于Transformer架构的轨迹控制参数预测模型,并设计平衡优化控制参数与物理轨迹的神经网络损失函数;最后,利用控制参数模型生成多条不同机动模式下的仿真轨迹数据,并输入轨迹预测模型以学习控制参数与轨迹数据的变化规律;此外,将测试轨迹数据输入训练所得的模型以获得模型性能的测试结果。结果表明:本文所提出的轨迹预测模型对不同机动模式下的高超声速飞行器轨迹都有良好的预测效果。With the increasing attack level of hypersonic vehicles,how to accurately predict their trajectory has become critical to defense research.This article has proposed a trajectory prediction model based on parameter estimation,which combines with the Transformer deep learning model.First,this paper developed a control parameter model that affects aerodynamic vehicle maneuvers and summarized the parameters’controlling rules under different maneuver modes.Secondly,this paper established a trajectory control parameter prediction model based on Transformer architecture and designed a neural network loss function that can balance the optimal control parameters with the physical trajectory.Finally,this article simulated trajectory data for multiple different manoeuvring modes and analyzed the changes in control parameters and trajectory data using the trajectory prediction model.This article has then obtained test results by inputting test trajectory data into the trained model.The results show that the trajectory prediction model proposed in this article performs precise prediction of hypersonic vehicle trajectories under different manoeuvring modes.

关 键 词：高超声速飞行器 轨迹生成 轨迹预测 空气动力学 Transformer模型 控制参数优化 

分 类 号：V448[航空宇航科学与技术—飞行器设计]