The Application In Eliminating Drift Error Of Inertial Navigation System Based On Pattern Recognition Algorithm With Improved Back Propagation

Autonomous Inertial Navigation System(INS) based on digital map is basis of realizing Terrain Aided Navigation(TAN). TAN is an effective approach which can eliminate or compensate excursion error of INS, It are also named after terrain matching. Actually, The system are combined with INS, wireless or radar height barometer and digital map. Which can be used for real-time removing accumulated error of INS with time. Therefore, The navigation precision could be improved so much, which is a notable advantage for it.Firstly, the digital map which is needed for implementing terrain matching is discussed. C++ programs are adopted to read corresponding height data which is given via three dimension terrain data file( Height .Data), at the same time, the three dimension terrain position data file(M file)was formed. Finally, all kinds of terrain pattern are produced by using Matlab simulation software, which are regarded as research foundation.Algorithm is the core of the terrain matching. This paper introduces briefly several comparatively mature algorithms and their principle, features and applications. there are TERCOM, SITAN ,similarity mature algorithm, the terrain entropy matching algorithm and mean absolute distance(MAD)algorithm and so on. Some common algorithms(methods) for enhancing navigation precision are also introduced in home and abroad. Such as: matching method in gravity aided INS , matching algorithm based on Bayes'Rule, sonar technology aided method, terrain aided navigation using fading memory and particle-filter algorithm. while an intelligent algorithm for erasing error of INS -The pattern recognition algorithm with BP neural networks is introduced in this paper. It has particular advantages in complex terrain field and owns speediness and validity for analyzing matching terrain position data in wide area.Even if in conditions of noise, the true position of carrier can be searched fast and accurately. Furthermore, improved BP neural networks have conquered several disadvantages existing before. Such as: the slow speed of convergence, local minimum and unsteady networks.At last, in order to validate the feasibility of the algorithm, the different terrain patterns can be simulated. It demonstrates that the algorithm achieves the desired matching effect. The study method in this paper also can be applied to the field of path programming. It can greatly improve the problem of path programming for all kinds of flight vehicles.