Gyroless satellite attitude determination using a SVD-Laplace Particle Filter

There is an increasing demand in the space industry for methods allowing for gyroless spacecraft attitude determination to either provide low cost sensor fusion or to recover from gyro failures. The gyroless configuration used in this paper is a combination of magnetometers and Sun sensors. Sun sensors are subject to measurement discontinuities during solar eclipses. Gyroless attitude sensing is known to increase sensitivity to measurement noise and model uncertainty. This paper presents a new Laplace Particle Filter with a Singular Value Decomposition to enhance attitude determination performance under these circumstances, while also preventing filter degeneracy. An orthogonal Procrustes problem is formulated to minimise the Wahba cost function and determine the attitude matrix from the Sun and magnetic field vectors and a third virtual predicted measurement vector, which is also available during Sun occlusion. The proposed filter is compared to a regularised particle filter approach, which handles particle filter degeneracy. Both filters are applied to the attitude determination of a small Earth observation satellite using three Sun sensors available during the sunlit phase and three orthogonal magnetometers. A Monte Carlo numerical simulation analysis demonstrates that the proposed particle filter significantly outperforms the regularised particle filter in terms of accuracy and robustness to the Sun sensor measurement discontinuities during solar eclipse and is also more accurate when both types of sensor measurements are available.