Multiple-model adaptive estimation for star identification with two stars

In this paper a multiple-model adaptive estimation approach is presented for star identification using only measurement data from two stars within the spacecraft sensor field-of- view. Star identification algorithms typically require four or more stars in the field-of-view to make the data association with known cataloged values robustly. When the minimum number of stars condition is not met, these algorithms are unable to converge to a solution. Sensors with a narrow field-of-view or observing a sparsely populated target region both complicate the star identification process with the current algorithms. This paper demonstrates the feasibility of a multiple-model adaptive estimation approach for robust star identification based on image measurements from two stars in the field-of-view. The concept utilizes successive observations of the same two stars and their inter-star angle along with an understanding of the sensor noise statistics. The measurements are com- pared to a set of candidate cataloged star pairs to create a residual which updates likelihood estimates for each possible star pair. Simulation results show robust convergence of the likelihood estimates to the true star pair when the measurements are corrupted with Gaussian noise. The compact and computationally efficient implementation is well suited for onboard spacecraft operation.