A new state estimator for spacecrafts at injection phase

In this paper, a smoother method was presented and employed for spacecraft's initial orbit determination. The smoother performs recursive forward filtering and backward smoothing based on the Bayesian filtering theory in the Gaussian domain. It avoids the linearization process which is indispensable in a least square estimator or an extended Kalman filter for astrodynamics. The smoothing style based on Rauch-Tung-Striebel smoother is shown to be optimal in statistical sense. The launch vehicle's GPS observation and ship-borne sensor's observation are used for the orbit determination of a launch vehicle's payload. The measurements include range, azimuth and elevation of a spacecraft. To evaluate and verify the idea of data fusion and performance of the proposed smoother, two simulations have been performed. The results of the first simulation show that state estimates utilizing combined measurements are much more precise than those based on only sensor's measurements. The results of second simulation show that the smoother is more robust and stable than the traditional batch least square estimator. The proposed method can be applied to long arc precise orbit determination or other nonlinear estimation problems.