Error modeling and monitoring methods of extended Kalman filter-based terrain referenced navigation system for integrity assurance

Abstract

Navigation safety is more crucial for terrain referenced navigation (TRN) systems used in middle-range unmanned aerial vehicles (UAVs) than those in cruise missiles flying through mountainous areas because middle-range UAVs may fly to regions and cities containing both combatants and civilians. For safety assurance of TRN systems, this thesis develops the integrity architecture for TRN systems and proposes a method to compute protection levels (PLs) to assure the required level of integrity. For obtaining PLs of TRN systems, all components of the TRN measurement errors must be identified and overbounded. This study models every error component: the radar altimeter error, the vertical error in the terrain elevation database, the lateral offset induced error, and the interpolation error. Based on the characteristics of TRN systems using terrain information for obtaining navigation solutions, this study develops new models for the lateral offset induced error and the interpolation error as a function of terrain roughness by examining the effect of terrain roughness on the error sources. The first simulation results demonstrate that the PLs obtained by applying the proposed error models can sufficiently overbound the navigation solution error for a TRN system under nominal conditions. The proposed method shows remarkable performance in regions with low terrain roughness. The integrity architecture of TRN systems is proposed including integrity monitoring based on the Extended Kalman Filter. This architecture consists of the integrity risk allocation, the nominal and fault hypotheses’ bounding and sensor faults monitoring methods. The PLs under nominal conditions applying the proposed error models are presented and the PLs under the fault hypotheses are calculated considering the implemented fault monitoring. The final PLs are determined using the PLs under every conditions. The second simulation results show the proposed final PLs can sufficiently overbound the navigation solution error for a TRN system under three hypotheses.