Two-stage robust extended Kalman filter in autonomous navigation for the powered descent phase of Mars EDL

This paper proposed a two-stage robust extended Kalman filter (TREKF) for state estimation of non-linear uncertain system with unknown inputs. In engineering practice, the extended Kalman filter (EKF) with unknown inputs of the non-linear uncertain system may be degraded or even diverged. The optimal two-stage EKF (TEKF) is designed to solve the unknown inputs. The robust EKF (REKF) is considered to solve the non-linear uncertain system for a long time. However, the information about the non-linear uncertain system with unknown inputs is always incorrect. To solve this problem, the TREKF is designed by using the advantages of the TEKF and REKF, furthermore, its stability is proved. Finally, the performances of the TREKF, which are compared with the results of the REKF, TEKF and EKF, are verified by illustrating a numerical example of the powered descent phase of Mars EDL (entry, descent and landing). These also verify that the unfavourable effects of the model uncertainties and the unknown inputs are reduced efficiently by using the TREKF for the miniature coherent altimeter and velocimeter and inertial measurement unit integrated navigation during the powered descent phase of Mars EDL.