A Marginalized Particle Filter approach to an integrated INS/TAP system

Abstract

Accurate and reliable navigation systems will become increasingly important in future aircraft applications, in particular within unmanned aerial vehicle systems. This paper describes a particle filter approach of integrating an Inertial navigation system (INS) with a terrain-aided positioning system (TAP) to achieve such a system. The integrated system is realized applying a marginalized particle filter (MPF) where the highly nonlinear TAP is designed tightly with the INS using one and the same filter. In order to better estimate the multi-modal errors in the altitude measurements, a first order Generalized Pseudo-Bayesian (GPB1) filter is used for this purpose. This will also reduce the number of particles in the MPF and therefore also reduce the computational workload. The performance of the algorithm has been evaluated using recorded flight data from the Saab Gripen fighter aircraft. Compared to an existing INS/TAP system based on a suboptimal integration of a point mass filter representing TAP and a single extended Kalman filter estimating the INS errors, the MPF approach is similar in performance but shows better results on convergence times when recovering after loss of data.