Research On Matching Area Selection Criteria For Gravity Gradient Aided Navigation

Inertial navigation system (INS) is the core to underwater navigation. However, theaccumulated drift errors of INS over time cannot be eliminated. The gravity gradientaided navigation system corrects INS errors regularly and gravity gradient aidednavigation system could improve is a kind of passive and autonomous navigation system.Gravity gradiometer, Gravity gradient reference map, adaptive area selection andmatching algorithm are the basic factors of Gravity gradient aided navigation systems. Inorder to improve the positioning accuracy of navigation systems by using the gravitygradient aided navigation methods, Gravity gradient adaptive area selecting is theprerequisite of it.In this thesis, extraction of Gravity gradient multi-features and selecting thecriterion of adapter area are researched. Firstly, Gravity gradient’s basic principle, thefeatures of the parameters and the normal matching algorithm are introduced. The bestmethod of searching the Gravity gradient multi-features are extraction. Then, the Gravitygradient matching area selection criteria are researched based on mathematical statistics,PCHP and support vector machine respectively these three methods. Among them, in themethod of mathematical statistics, large of data were analyzed at the beginning with theactual navigation and positioning experiment, and the standard deviation, the energy andthe absolute roughness are used as constraint conditions of gravity gradient matchingarea directly. In the method of PCHP, nine features were analyzed, and most of thefeatures information was retained. In method of SVM, training set sample label of SVMwere identified, based on the conclusion of mathematical statistics. Adaption area isdivided through the study of the gravity gradient data. Three methods have their owncharacteristics and application.The gravity gradient aided positioning simulation resultsshow that the effect of the matching navigation in the adaptation area is markedlysuperior to the effect in the non-adaption area, the position error is less than a grid, andmatching rate is greater than90%. The selection criteria based on SVM has obviousadvantages from the intelligence, speed and accuracy. At last, summary and prospect aregiven.