基于数据驱动的非合作航天器姿态估计与预测方法

针对临近操作对非合作航天器的相对导航问题，考虑角速度测量缺失以及视觉特征丢失，提出了一种融合乘性扩展卡尔曼滤波和姿态预测的框架，实现了对非合作航天器的姿态估计和预测。采用惯性参数对状态向量进行了扩维，在缺少角速度测量的情况下预估了非合作航天器的姿态和相对转动惯量的比值。基于函数拟合和神经网络分别设计了两种姿态预测方法，解决了传统方法误差随时间累积的问题，有效减少了计算成本。最后，通过数值仿真验证了滤波估计和姿态预测算法的实时性和准确性。

Data-Driven-Based Attitude Estimation and Prediction for Noncooperative Spacecraft

In this work, a novel solution to the problem of the relative navigation during the proximity process is proposed to realize the attitude estimation and prediction of noncooperative spacecraft. It is a framework incorporating multiplicative extended Kalman filtering and attitude prediction. And the lack of angular velocity measurement and the loss of visual features are considered. By employing the augmented state vector, including inertial parameters, the attitude and relative ratios of the moment of inertia for noncooperative spacecraft are estimated in the absence of angular velocity measurement. Two attitude prediction methods, function fitting and neural network, are presented respectively. The problem of accumulating error over time of traditional methods is improved and the computational cost is reduced. The effectiveness and real-time performance are validated via numerical simulations of the proposed attitude estimation and prediction algorithms.