基于SRUKF的偏心仅测角相对导航方法

针对以单个光学/红外相机作为相对导航敏感器的航天器，提出采用相机偏心安装和平方根无损卡尔曼滤波（SRUKF）两种方法，解决自主交会时中远距离和近距离全过程的仅测角导航(AON)问题。在观测相机安装时，人为设置一个分米量级的安装偏心，按照观测距离分段建立相机观测模型并推导了观测敏感矩阵，从理论上证明了径向和法向相机安装偏心可以改善近距离v-bar相对位置保持点的观测度，并且对远距离导航精度没影响。构建了基于SRUKF的仅测角导航方法，该方法不需要强制要求初始估计误差满足小量假设和零均值高斯分布假设。经算例验证，在远距离时，采用本文提出的仅测角导航方法，其导航精度比传统扩展卡尔曼滤波（EKF）方法提高了一倍，计算负载降低了约10%；在近距离时，解决了仅测角导航v-bar相对位置保持点不可观测问题；并且当初始估计误差呈非高斯特性时，依然可以实现对相对距离的无偏估计。

Square Root Unscented Kalman Filter Based Angles Only Relative Navigation Using Camera Offset

A square root unscented Kalman filter (SRUKF)-based angles-only navigation (AON) method is proposed for spacecraft rendezvous using single optical/infrared camera as relative navigation sensors both in far/middle and close-in range. To address the well-documented range observability problem of AON, a decimeter scale camera installation offset from the mass center of the chaser is intentionally set. A camera measure model is constructed by the range scope and the observation sensitivity matrices are furtherly derived. We theoretically prove the component in cross-track or radial generates range observability for v-bar relative station keeping. Small error and zero mean Gaussian distribution hypotheses are not essential any more using SRUKF. Through numerical simulation, in far/middle range, the estimation error of the SRUKF-based AON is just half of the traditional extended Kalman filter (EKF)-based AON. In close-in range, the range observability problem is solved when the camera offset is considered. SRUKF and EKF have the same order computational burden for the state estimation, but SRUKF is about 10% faster than the EKF due to using the two powerful linear algebra techniques, QR decomposition, Cholesky factor updating. Last but not least, even with the non-Gaussian initial estimation error, the unbiased estimation of the relative range is still achieved.