Research On Error Suppression Technology Of Intelligent Integrated Navigation

The complementary advantages of SINS and GNSS have become the mainstream of integrated navigation,which is widely used in military and civil fields.At present,MEMS inertial components are widely used in vehicle-mounted integrated navigation system,which reduces the cost of vehicle-mounted integrated navigation system.However,MEMS inertial Measurement Unit(MIMU)is the core measurement component of SINS,and its output measurement signal contains large random noise and low measurement accuracy,which is an important factor leading to divergence of SINS navigation error.In tunnel and other environments,GNSS cannot guarantee the stability of the output signal.When GNSS failure happens,SINS works independently and the system navigation error rapidly diverges.To solve the above problems,this paper takes low-cost SINS/GNSS integrated navigation system as the research object and aims to suppress the divergence of navigation errors of integrated navigation system in the case of GNSS failure.The following research works are carried out:(1)Firstly,the navigation principles of SINS and GNSS are studied,and the errors of the two subsystems are analyzed.The structure and mathematical model of the integrated navigation system are established,which provides theoretical support for the subsequent design of error suppression algorithms.(2)Aiming at the problem that the influence of MIMU random error on the prediction accuracy of GNSS navigation results is ignored in existing AI-aided SINS navigation schemes,the error characteristics of MIMU are analyzed,and the error sources and error characteristics of MIMU are revealed to be different from those of GNSS.In order to suppress the random error of MIMU measurement output and improve the accuracy of artificial intelligence model training samples,a joint denoising algorithm based on direct fast iterative filtering(DFIF)and wavelet threshold denoising method was proposed.DFIF was introduced to improve the inherent defect of lacking mathematical foundation of EMD,and the problem of noise modal differentiation after DFIF decomposition signal was completed was solved by constructing a distinction index integrating linear and nonlinear relations.The superiority of DFIF in modal identification ability was further verified by the non-stationary signal simulation noise reduction experiment,and the effectiveness of the joint noise reduction algorithm was proved.(3)To solve the problem of Kalman Filter(KF)failure and rapid divergence of SINS navigation errors in the case of GNSS failure,Extreme Learning Machine(ELM)was used to predict the GNSS navigation output and correct SINS navigation errors.The shortcomings of Yin-Yang pair optimization(YYPO)algorithm were analyzed,and an adaptive speed regulating factor was proposed to enhance the global convergence of YYPO algorithm,and the improved YYPO algorithm was applied to the optimization of output threshold of hidden layer of ELM,which improved the time series prediction performance of ELM.(4)Finally,the overall scheme of intelligent integrated navigation error suppression under GNSS failure is designed based on the algorithm proposed in this paper.The real output data of MIMU is collected in the laboratory environment.Taking MEMS gyroscope as an example,the static MEMS gyroscope output and dynamic MEMS gyroscope output are denoised and verified respectively.The results show that the joint denoising algorithm has better denoising performance than the traditional denoising algorithms.Through open source sports car data,the error suppression scheme is verified,and the navigation error suppression of low-cost vehicle-mounted integrated navigation system in the case of GNSS failure is realized.