Nanosatellite attitude estimation, in-orbit sensor calibration and fault detection using kinematic relations only

Abstract

This study focuses on estimating the attitude angles of a nanosatellite using vector measurements from onboard magnetometers and sun sensors. The estimation scheme employs a nontraditional filtering approach that relies solely on a kinematic model propagated with gyroscope and magnetometer data. The overall process consists of two main stages. In the first stage, attitude measurements are obtained using a single frame method based on singular value decomposition (SVD). In the second stage, these measurements are fused within an Extended Kalman Filter (EKF) that incorporates a linearized attitude model together with nonlinear magnetometer observations. Integrating these two stages yields a unified estimation scheme, referred to as the SVD Aided EKF, which enables accurate, simultaneous estimation of attitude angles as well as gyroscope and magnetometer biases. To detect attitude sensor faults, this study proposes fault detection statistics based on the central Wishart matrix whose diagonal elements are obtained from the components of the innovation vector. The proposed innovation-based fault detection algorithm was applied to the SVD-Aided EKF attitude estimation system to detect nanosatellite attitude sensor faults.