Junction-tree-based cooperative orbit determination for crosslink-augmented satellite constellations

Abstract

The lack of ground-tracking resources has become a primary bottleneck for the Chinese BeiDou navigation satellite system. As crosslinks have been widely recognized as a promising augmentation for the autonomous navigation of the global navigation satellite system, this article studies a new decentralized data fusion method for orbit determination of crosslink-augmented satellite constellations. In the new solution, the system is modeled as a probabilistic graphical model, the dynamical Bayesian network, and a graphical method named junction tree is introduced to analyze the structure of the system. By dividing and arranging the predictions and estimations of junction tree in a proper sequence, a new decentralized solution with centralized equivalent precision is designed. As the solution requires satellites in the constellation cooperating with each other through communications and measurements, it is named junction-tree-based cooperative orbit determination. Simulation results indicate that junction-tree-based cooperative orbit determination has centralized equivalent precision and good robustness after satellite failures, whereas centralized solutions such as extended Kalman filter may suffer from processing center malfunctions. Junction-tree-based cooperative orbit determination could not only serve as an optional choice for global navigation satellite system autonomous navigation but also be used as a general scheme for decentralized data fusion in cooperative systems such as unmanned aerial vehicle formations and so on.