Processing Method Of Orientation Error For Bionic Polarization Compass And INS Combination

This study develops a orientation error method aiming at high-accuracy and strong-robust orientation of the bionic polarization compass/inertial navigation system(PC/INS)for unmanned aerial vehicle(UAV)when GPS signal is rejected and INS error is accumulated easily with time,even though PC signal is unavailable in complexible environments.The critical methods not only contain the analysis and processing of orientation error for the compass,but also the seamless integrated orientation menthod for PC/INS.First,we emphasize a multiscale transform(MST)denoising method for the bionic polarized light compass.Subsequently,an attitude-changed error modelling and compentation method with gate recurrent unit(GRU)neural networks,and a seamless integration method of PC/INS output data with cubature Kalman filter(CKF)using a self-learning multi-rate residual correction algorithm are propsed for improving the compass oreintation accuracy in various ways.Finally,the static,rotational,and UAV fight experiments are carried out to demonstrate the effectiveness of the proposed schemes.The entire research results of this dissertation are summarized as follows.(1)Propose an Ao P image denoising method and a heading data denoising method based on multiscale transform(MST).First,a multiscale principal component analysis method(MSPCA)utilizing one-dimensional image entropy as classification criterion is directly implemented to suppress the noise in the acquired polarization image.Subsequently,a principal component analysis(PCA)is directly employed to suppress the noise in an Ao P image,in which two-dimensional empirical mode decomposition(BEMD)is utilized to decompose the original noisy Ao P image into multiple image IMFs.The Aop image denoising method with PCA based on a principal component contribution rate is introduced to denoise each scale IMF including solar meridian information.Finally,a multiscale time-frequency peak filtering method(MSTFPF)is exploited for removing the noise in the heading data output from a polarized light compass,in which the output data is decomposed into distinct IMFs by ensemble empirical mode decomposition(EEMD).These two approaches are combined to significantly reduce the heading error affected by different types of noises.Our experimental results indicate that the proposed multiscale transform denoising method exhibits high performance for the compass orientation accuracy compared to the existing works.The static orientation error is 0.0735 °(RMSE),turntable orientation error is 1.2365 °(RMSE),and UAV orientation error is 0.3116 °(RMSE),respectively.(2)Propose an orientation error modelling and compensation scheme with GRU deeplearning neural networks.We first introduce a comprehensive analysis of heading error in terms of variable attitude angles of the compass including the angle between the solar meridian and the body axis of a carrier(A-SMBA),the pitch angle and the roll angle.A novel heading error modeling and compensation method for attitude-changed of the polarization compass by GRU deep-learning neural network is developed subsequently.The different dynamic experiment results demonstrate the proposed heading error modeling and compensation method performs best contrast to state-of-the-art algorithms in predicting the dynamic orientation accuracy by different actual experiments.The turntable orientation error is 0.4560 °(RMSE)and the UAV airborne orientation error is 0.5218 °(RMSE),respectively.(3)Propose a seamless integration method of PC/INS data with cubature Kalman filter(CKF)using a self-learning multi-rate residual correction algorithm.A seamless integration model based on CKF-MR is first created for PC/INS orientation system.Secondly,when PC works normally,an algorithm of PC/INS data employing CKF fusion based on muti-rate residual correction(CKF-MRC)is developed to revolve the problem that the output data frequency of PC and INS is inconsistent and the accuracy of the whole orientation system needs to be further improved.Finally,a deep self-learning seamless integrated orientation method on the basis of LSTM is proposed to predict the heading data output from PC when the PC cannot work normally due to complex environmental conditions such as occlusion.The experimental results demonstrate that the seamless integrated orientation method suggested in this paper not only effectively improves the PC/INS data output frequency(100HZ)and orientation accuracy(RMSE:0.21°)of the integrated orientation system when the PC is not affected,but also maintains high orientation accuracy(RMSE:0.53°)when the PC is not available for a short time due to the influence of the environment.