Digital Contour Model Based Terrain Referenced Navigation

Abstract

The importance of accurate navigation cannot be overemphasized. The majority of the various types of navigation systems are categorized into inertial navigation system (INS) and Global Positioning System (GPS). Because of their complementary features, INS/GPS integrated navigation has been developed as the principal navigation solution. In spite of the distinct advantages of a GPS, it is still difficult to entirely rely on an INS/GPS integrated system due to the dependency on external satellite signals and vulnerability to intentional interference. Terrain referenced navigation (TRN) has been regarded as an alternative choice to GPS for the purpose of dependable stand-alone navigation. TRN had been developed for a long time ago in batch or sequential manners, but the development has been slow or stagnated because the technology at the time was limited to fully implement the algorithm at its best. The constraints of TRN development was real-time computing power, data-storage capability, accuracy of terrain sensors, precision of database, etc. Modern advanced technology not only frees TRN from these limitations, but it also enables new TRN methods. The requirement for stand-alone navigation that is not affected by the external radio environment, and the development of modern advanced technology that can overcome the limitations of the past TRN have made it possible to launch new terrain referenced navigation. The novel concept that is noted in this study is measurement area. When a terrain height is given by altimeter measurements, it is important to find the same altitude contour line through the terrain database and to find the measurement area, which is a bundle of contour lines, taking into account the altimeter error. Measurement area is an area likely to contain the true position, and it is possible to develop the algorithm to acquisition mode or track mode according to the assumption of the probability distribution of this area. Assuming that the probability of measurement area is uniformly distributed, and introducing the positioning technique by proximity, the true position can be obtained by intersection of successive measurement areas. The contour intersection-based acquisition method has the advantages of very fast computation. However, it shows the problem that the intersection area becomes an empty set due to erroneous measurement, an inaccurate database, or an imperfect measurement area. The contour collection-based acquisition method has been proposed to solve the problem of empty set in intersection method. Collection method shows that stable and smooth position results can be obtained without resetting until the time when the initial INS relative error is well maintained. However, this method requires a finer division of the grid, so the amount of computation is relatively large. Also, due to the influence of the INS error, the acquisition error is grow with time. In order to prevent the divergence of the acquisition error, a filter must be considered to compensate the error of the INS. The tracking method using a point mass filter can be implemented by weighting the contours constituting the measurement area. The contour weight-based tracking method can operate reliably without divergence if the initial area assuming the probability density function is smaller than a certain size. In order to create a new terrain referenced navigation, this thesis proposes a contour based terrain referenced navigation combining the acquisition methods and tracking method described above. Without knowledge of the initial position error, the intersection method quickly removes the meaningless area, and the collection method complements the intersection method to keep the acquisition robustly for meaningful areas. After that, the INS error is compensated by the tracking when converging enough to avoid divergence risk. This process consists of a coarse acquisition mode, a fine acquisition mode, and a track mode and are integrated into one algorithm in the contour based terrain referenced navigation. Through Monte Carlo simulations, we confirmed that the proposed terrain referenced navigation works well regardless of terrain shape. As a result of computer simulations for various initial position errors and terrain, the acquisition results show a sufficient positioning accuracy for tracking, and the tracking results based on the acquired outputs show excellent performance over the entire tracking period. In this thesis, we can see that the proposed method is a new terrain referenced navigation that overcomes the disadvantages of batch processing and sequential processing.