Measurement Selection for Autonomous Satellite Constellation Navigation Using Parallel Extended Kalman Filters

Abstract

The autonomous navigation method for a constellation with several Earth satellites and a lunar satellite using inter-satellite range measurements is studied. The presented PEKF algorithm is suitable for autonomous satellite constellation navigation. As the inter-satellite range measurements are susceptible to the exterior disturbance, this paper focuses on the measurement selection problem with noise statistic uncertainty. To select the appropriate measurements adaptively, a parallel extended Kalman filters (PEKF) is presented, where each extended Kalman filter (EKF) is designed to process a subset of the measurements, and the estimation results of the parallel filters are combined on the basis of the individual residual sequences. The derivation of the Cramer-Rao lower bound (CRLB) shows that introducing more measurements could leads to an improvement in the theoretical filtering performance. The performance advantage is illustrated in comparison with the EKF and the traditional multiple-model adaptive estimation (MMAE). An effective PEKF algorithm is developed to deal with the measurement selection problem for an uncertain system, such that the available measurements which have the potential to improve the estimation accuracy are utilized appropriately.