基于深度神经网络的最小星间距离计算方法  

A DNN-based method for calculating minimal inter-satellite distances

作　　者：王柯鑫 党朝辉 Wang Kexin;Dang Zhaohui(School of Astronautics,Northwestern Polytechnical University,Xi'an,710072,China;National Key Laboratory of Aerospace Flight Dynamics,Northwestern Polytechnical University,Xi'an,710072,China)

机构地区：[1]西北工业大学航天学院,西安710072 [2]西北工业大学航天学院航天飞行动力学技术重点实验室,西安710072

出　　处：《空天技术》2024年第4期76-84,共9页Aerospace Technology

摘　　要：针对最小星间距离求解方法的普遍效率低及适用性较弱等缺点,提出了一种在求解一对处于圆轨道进行非周期共面相对运动的卫星最小星间距离时,兼顾适用性、精度和计算效率的方法。该方法利用深度神经网络(Deep Neural Networks,DNN)的优势来拟合两颗卫星的初始相对状态与最小星间距离之间的关系。所提出的方法能够仅通过任何给定初始时间的位置和速度信息来计算最小距离。在分析并推导了星间距离函数最小值点范围的基础上,设计了数据集的构造方法。建立了训练DNN的模型,并重复进行训练过程,直至达到收敛。通过不同的仿真实例验证了DNN的训练结果。结果表明,基于DNN的计算方法可以几乎实时地计算最小距离,相应的相对误差几乎低于0.2%。Aiming at the shortcomings of the minimal inter-satellite distance solving methods,such as the general low efficiency and weak applicability,a method that can balance the applicability,accuracy and computational efficiency when solving the minimal inter-satellite distance for non-periodic coplanar relative motion of a pair of satellites is proposed.The principal idea of this method is utilizing the advantage of Deep Neural Networks(DNN)to fit the relationship between the initial relative state of two satellites and the minimum inter-satellite distance.The proposed method is able to calculate the minimal distance only through the position and velocity information at any given initial time.The close-range relative motion equations between satellites are established,based on which the analytic inter-satellite distance function is obtained.Through the analysis and derivation of the function,the range of the minimal inter-satellite distance with time is analytically obtained.The construction method of the dataset,which maps the initial states and the minimal distances,is designed to obtain a great number of mappings.Then a model to train the DNN is established and the training process is conducted repeatedly until the convergence is achieved.The training results of the DNN are verified through different simulation examples.The results demonstrate that the DNN-based calculation method can compute the minimal distance nearly in real-time and the corresponding relative errors are almost below 0.2%.

关 键 词：圆轨道 非周期相对运动 共面 最小星间距离 深度神经网络 

分 类 号：V19[航空宇航科学与技术—人机与环境工程]