Decentralized Cooperative Orbit Determination For Satellite Constellations

Some space missions require persistent orbit determination of satellite constellations,e.g.the Global Navigation Satellite System(GNSS).However,this requirement is usually based on establishing ground tracking stations worldwide,which is not practicable for every country,e.g.the Chinese BeiDou navigation satellite System(BDS).With recent progresses in the technology of inter-satellite measurement and communication,also known as satellite crosslinks,a new solution arises: orbit determination with the aid of crosslinks.Aimed at autonomous operation of the BDS,this thesis studies the following four problems.(1)Since orbit determination for satellite constellations based on inter-satellite crosslinks may be faced with the problem of constellation rotation,a cooperative orbit determination method based on crosslinks and ground anchor stations is introduced.State space models such as the dynamic model and measurement model are introduced.And the observability of the system is analysed by utilizing the local observability theory.(2)Based on the state space model,three new decentralized cooperative orbit determination algorithms based on the Extended Kalman Filter(EKF)is introduced,which are the Full Order Extended Kalman Filter(FOEKF),the Reduce Order Extended Kalman Filter(ROEKF),and the Iterative Cascade Extended Kalman Filter(ICEKF).All the algorithms are evaluated from the viewpoint of precision,computational load,and communicational load.(3)Based on the probabilistic graphical model,a new decentralized cooperative orbit determination algorithm based on the Junction Tree(JT)is introduced.The new algorithm is equivalent to centralized Bayesian Estimation(BE).For space applications based on computers with limited machine word length,the Square-Root Junction Tree(SRJT)algorithm is developed.SRJT has better numerical stability than BE and JT.(4)Aimed at the problem of autonomous orbit determination of BDS,a modified orbital elements based on the broad ephemeris and the related partial derivatives for orbit determination are introduced.Aimed at the problems for current algorithms caused by ill-conditioned matrices,a new centralized algorithm named the Windowed Iterative Square-Root Extended Kalman Filter(WISREKF)and a new decentralized algrithm named the Windowed Iterative Square-Root Junction Tree(WISRJT)are introduced.The algorithms of SRJT and WISRJT may serve as supplements to the decentralized data fusion method based on precise BE,while WISREKF as supplements to the centralized data fusion method based on EKF.The modified orbital elements based on broadcast ephemeris and the partial derivatives for orbit determination may serve as references to autonomous ephemeris fit for GNSSs.The decentralized algorithms may serve as general schemes for decentralized data fusion in other cooperative systems.