Research On Improved Model Of Inertial Navigation Aided Seamless Positioning

Inertial navigation system is a kind of passive navigation equipment with strongautonomy, high precision in short period, and possesses continuous navigationinformation output ability compared with other navigation systems, therefore, it hasgreat application value in the military and civilian. This paper focuses on theimproved model of inertial navigation aided seamless positioning, which mainlycovers IMU random error identification, anomaly detection and the improved robustand adaptive filtering model, machine learning aided shielding zone intelligentnavigation algorithm, IMU aided pedestrian track help reckoning with zero velocityupdate indoor navigation and positioning, the main work and contributions aresummarized as follows:(1) In order to conquer the problems that the amount of computation of theconventional Allan variance is huge, and the coefficient matrix cannot be modifiedwhen random error parameters was fitted by the least square model, this paper putsforward an Allan variance simplified estimation algorithm based on the WTLS.Through the data analysis, the results showed that, the algorithm can greatly reducethe calculation amount, accelerate the speed of operation and maintain the accuracyof result.(2) Navigation performance of GPS/INS system will be drastically reducedwhen GPS outages. With Sage adaptive filtering, an improved RBF neural networkaided navigation is proposed. GPS pseudo-positon is predicted with the nearestneighbor clustering algorithm based on genetic algorithm parameter optimization,which can be used to assist navigation solution. The study provided intelligentauxiliary the navigation algorithm for short-term reliability for outdoor shaded area.(3) An improved nonlinear filtering model is put forward, which canindependently judge ill-matrix to select different robust strategy. And a robustadaptive Kalman filtering algorithm is put forward based on genetic algorithm andsupport vector regression. The algorithm addresses the limits of anomaly detectionon condition of lacking redundant observations. which choose robust or adaptiveKalman filtering for purpose of adjusting contribution of observations and dynamicalmodel to the results. The algorithm improves the reliability and accuracy ofnavigation solutions.(4) A LS-SVR aided adaptive unscented Kalman filter (UKF) algorithm withmultiple fading factors based on singular value decomposition (SVD) is proposed. The algorithm uses the LS-SVR to weaken the influence of the observedabnormalities on the residual vectors. Singular value decomposition instead ofunscented transformation is adopted to suppress negative definite variation in prioricovariance matrix of UKF. The algorithm provides an advanced filtering formultivariable complex system(5) Based on the time-frequency transform analysis of noise characteristics ofinertial sensors used in indoor positioning. A pedestrians dead reckoning algorithm isproposed with pedestrian detection based on FIR filter With the generalizedlikelihood ratio test used to identify zero veloity, adaptive filter based on reliableobservations is used for ZUPT navigation model. The algorithm can weaken thenoise in the observed value of information, improve the directional stability, enhancethe reliability of the zero speed detection, improve the inertial aided indoornavigation accuracy.