COMPARISON OF MODULAR AND CENTRAL TERRAIN REFERENCED NAVIGATION FILTERS

摘要

The measurement equation of a terrain referenced navigation (TRN) system is non-linear because of the influence of a reference map connecting position and terrain height. It is possible to design different navigation filters that process radar altimeter measurements based on this measurement equation. The first proposed system is based on an extended Kalman, the second on a sigma-point, the third on a bootstrap, and the fourth on a modified particle filter. The first and second filter approximate the state probability density function by a Gaussian distribution, while the third and fourth filter allow a more general density estimation by particles. All four systems can be used as central navigation systems processing radar height measurements and estimating position, velocity, and attitude. Especially for both particle filters the system performance is not ideal, because Of the weak observability of the velocity and attitude states. In another situation where the TRN-system has to be separated from the central navigation a setup with a central Kalman filter and a terrain referenced navigation module with a reduced state space is proposed. Compared to the central TRN filters the overall accuracy of the modular filter is reduced due to the separation of the non-linear TRN-module and the linear central navigation. Anyhow, because of the reduction of the state space the modular particle filters can gain robustness and achieve similar accuracy than the central systems.