Improving extended Kalman filter algorithm in satellite autonomous navigation

With the goal of reducing dependence on ground tracking systems, satellite autonomous navigation technologies are developed quickly in the recent several decades. However, precise orbit determination at high orbital altitudes is an important and challenging problem. In this paper, the nonlinear real-time orbit determination problem is investigated. Combined with satellite dynamical model, extended Kalman filter is explored to estimate satellite orbit parameters. Further, considering errors occur in linearization processing, two improvements for the extended Kalman filter algorithm, i.e. extended Kalman filter-I and extended Kalman filter-II, are proposed based on Lagrange's mean value theorem, and respectively focus on choosing better linear expansion point and Jacobian matrix calculation point. Extensive simulations show that extended Kalman filter-I and extended Kalman filter-II significantly enhance orbit accuracy, compared with extended Kalman filter. And the increases in calculation complexity are acceptable. Finally, the robustness of extended Kalman filter-I and extended Kalman filter-II is analyzed by given different initial position errors, and results show that extended Kalman filter-I and extended Kalman filter-II have better robustness than extended Kalman filter.