Adaptive filtering applications to satellite navigation

Abstract

Differential Global Navigation Satellite Systems employ the extended Kalman filter to estimate the reference position error. High accuracy integrated navigation systems have the ability to mix traditional inertial sensor outputs with navigation satellite based position information and can be used to develop high accuracy landing systems for aircraft. This thesis considers a host of estimation problems associated with aircraft navigation systems that currently rely on the extended Kalman filter and proposes to use a nonlinear estimation algorithm, the unscented Kalman filter (UKF) that does not rely on Jacobian linearisation. The objective is to develop high accuracy positioning algorithms to facilitate the use of GNSS or DGNSS for aircraft landing. Firstly, the position error in a typical satellite navigation problem depends on the accuracy of the orbital ephemeris. The thesis presents results for the prediction of the orbital ephemeris from a customised navigation satellite receiver's data message. The SDP4/SDP8 algorithms and suitable noise models are used to establish the measured data. Secondly, the differential station common mode position error not including the contribution due to errors in the ephemeris is usually estimated by employing an EKF. The thesis then considers the application of the UKF to the mixing problem, so as to facilitate the mixing of measurements made by either a GNSS or a DGNSS and a variety of low cost or high-precision INS sensors. Precise, adaptive UKFs and a suitable nonlinear propagation method are used to estimate the orbit ephemeris and the differential position and the navigation filter mixing errors. The results indicate the method is particularly suitable for estimating the orbit ephemeris of navigation satellites and the differential position and navigation filter mixing errors, thus facilitating interoperable DGNSS operation for aircraft landing.