The research of satellite attitude determination algorithm based on federal filter

In order to obtain satellite attitude with better precision and reliability by using a multi-sensor satellite attitude determination system with uncertain model error, federated filter structure is adopted. The subfilter of federated filter is an improved Nonlinear filter, which is suitable for estimation of nonlinear system with uncertain model error by combining the Predictive filter and second-order divided Difference Filter (NPDF). The proposed method estimate model error and system state estimation by fusion multi-sensor information. The simulation results show that the federated filter with feedback, compared with no feedback, obtained better precision that is similar with that of centralized filter. 5 kinds of information-sharing methods are compared, Frobenius norm of estimated error variance matrix algorithm and variance matrix eigenvalue trace algorithm obtained relatively better estimation precision.