Research On Precision Evaluation Method Of Inertial Navigation System Based On Least Squares Support Vector Machine

Inertial navigation system is a concealed navigation system immune to external interference and widely used in aerospace,aviation and marine and other fields.It is an important safeguard for weapon systems like missile to realize quick and precision strike.Inertial navigation system has a complex structure with cross-coupling between error sources,which make it difficult to carry out intuitive evaluation on its operation accuracy through error parameters of inertial navigation systems.At the same time,the battlefield situation changes fast and combatants need to conduct accuracy evaluation and tactic planning for weapon systems in a short time,correctly and quickly evaluate the navigation accuracy of inertial navigation systems and provide quick and effective accuracy estimates of inertial navigation systems for tactic deployment.In this paper,we carried out research on the foresaid issues:We analyzed the operating principle of inertial navigation systems,particularly deduced the error equation of wander azimuth inertial navigation systems,found out analytical solutions for error equations of inertial navigation systems using the characteristic root method and analyzed the impact of various error sources on navigation accuracy and its error propagation characteristics,as well as the main factors that influence the navigation accuracy of inertial navigation systems to provide a theoretical basis for subsequent researches on evaluation methods for navigation accuracy.The effects of heading effect error and accelerometer quadratic error on the navigation accuracy of INS are studied in order to get a more perfect error model of INS and improve the accuracy of navigation solution.The error model with these two error sources is more accurate than the original error model.Based on the solution of the improved error model,an estimation method of navigation accuracy of inertial navigation system is proposed,and an evaluation software based on error model calculated by INS navigation accuracy evaluation method is designed and implemented.In order to improve the rapidness of the evaluation method,a dual-strategy adaptive pruning least squares support vector machine(LSSVM)navigation accuracy evaluation method is proposed.The improved pruning algorithm uses the global representative point selection method to determine the initial working sample set.The pruning strategy is realized by the improved method of removing unless the key sample points,and the verification mechanism of the reduced learning model is designed.Compared with the standard least squares support vector machine,the evaluation accuracy is kept in the same order of magnitude and the prediction time is reduced by half.Aiming at the problem that the data dimension of inertial navigation system is large and the training time of evaluation model is too long due to the dimension disaster occurred in the training process of least squares support vector machine.A navigation accuracy evaluation method which combines the principal component analysis with the double strategy adaptive pruning least squares support vector machine is proposed.Compared with the standard least squares support vector machine model,the training time of this evaluation method is reduced by an order of magnitude,which achieves the purpose of accurate and fast estimation of navigation accuracy of INS.