Study on attitude measurement algorithm of star sensor

For the problem of low accuracy of traditional vector fixation method, the mainstreammulti-vector astronomical attitude determination algorithm based on star sensors is studied, and singular value decomposition(SVD) algorithm, double vector fixed attitude algorithm, least square(LS) algorithm, quaternion estimator(QUEST) algorithm and linear estimation algorithm are derived theoretically in detail. The linear estimation algorithm disperses the coordinate transformation equations of each vector pair and deduces the corresponding quaternion form, then combines the coordinate transformation equations of all vectors, and linearizes the quadratic quaternion matrix equation based on the M-P pseudo-inverse operation of rectangular matrix. On this basis, considering the influence of random noise, a robust operation is performed on the quaternion matrix equation with degree 1.The method has good innovation and practicability. With the goal of minimizing the Wahba loss function, the performance of the attitude determination algorithm is compared and analyzed. The simulation and experimental results show that the linear estimation algorithm has faster solving speed and higher accuracy than the other four algorithms.

This article provides a detailed theoretical derivation of the star sensor attitude determination SVD algorithm, dual vector attitude determination algorithm, LS algorithm, QUEST algorithm, and linear estimation algorithm. Among them, the linear estimation algorithm fully utilizes the properties of pseudo inverse matrices to reduce the degree of quadratic quaternion matrix equations, effectively establishing a linear theory to solve the Wahba problem. Establish an objective optimization function and use the coordinate information of the starlight vector in the celestial and star sensitive systems to simulate and compare the pose determination performance of several different algorithms. Theoretical analysis and simulation experimental results show that the linear estimation algorithm, SVD, and QUEST algorithms have the highest pose determination accuracy, and the linear estimation algorithm has the highest computational efficiency, indicating that the comprehensive performance of the linear estimation algorithm is the best.