Bayesian bootstrap filtering for the LEO satellite orbit determination

This paper presents a new Bayesian bootstrap filtering approach for the low Earth orbit (LEO) satellite orbit determination using magnetometer measurements. To overcome the divergence problem of the extended Kalman filter (EKF), a hybrid bootstrap filter is proposed. The filters uses the magnitude of the magnetic field vector as measurements, and does not need the attitude information of the satellite. According to our simulation study our algorithm converges rapidly to the arc orbital parameters even in the presence of large initial position errors, and attains a few tens of kilometers of the position estimation error