INS/GNSS Integrated Navigation Using Artificial Neural Network

Since the 21st century,with the rapid development of the inertial navigation system(INS),its application has become more and more extensive,especially the integrated navigation with the global satellite navigation and positioning system(GNSS)is regarded as reliable,stable,The most complementary navigation system,especially to adapt to the shortcomings of inertial navigation system errors accumulated over time and the characteristics of satellite navigation systems that need to receive electromagnetic signals in real time.However,when the carrier is running near tall buildings in the city,in tunnels or canyons,the electromagnetic signal of satellite navigation will be affected at this time,which will cause the positioning accuracy of the navigation system to decline.Therefore,it is of great significance to study the algorithm based on the INS/GNSS integrated navigation data when the satellite navigation signal is interrupted,so as to improve the positioning accuracy of the integrated navigation.Under normal circumstances,There are two main methods to improve the positioning accuracy of INS/GNSS combined navigation systems:first,by improving the accuracy of the hardware component,such as improving the accuracy of the gyroscope and accelerometer,but the improvement of the hardware often requires high costs;the other is Improve the method of data fusion to improve the accuracy of navigation and positioning.The algorithms introduced in this article mainly include:Kalman Filter(KF),Extended Kalman Filter(EKF),Recurrent Neural Network(RNN),Long Short-term Memory Neural Network(Long ShortTerm Memory,LSTM).In order to overcome the short-term obstruction of satellite signals that affect the positioning accuracy of the integrated navigation,the RNN’s INS/GNSS integrated navigation algorithm is used,and after subsequent analysis,the RNN improved network LSTM is applied to the INS/GNSS integrated navigation.First,it introduces the main issues to be addressed in this topic and the research context,briefly summarizes the basic theory of INS/GNSS integrated navigation,including the definition of coordinate transformation in integrated navigation system and the analysis of common methods of INS/GNSS integrated navigation system.The state equation and measurement equation of the system are established,verified and analyzed.In the SINS part,the calculation of attitude,speed and position,error analysis,and the establishment of a simulation platform are given;in the GNSS part,the basic positioning principle and error analysis are given.Secondly,Basic models and learning algorithms for artificial neural networks are presented,as well as the data fusion algorithms KF and EKF,which are commonly used in combined GNSS/SINS navigation.A linear model based on KF and a nonlinear model based on EKF are constructed to lay a foundation for subsequent comparison and verification.Base.Aiming at the problem of GNSS signal failure,an INS/GNSS integrated navigation scheme based on Recurrent Neural Network(RNN)is proposed.The program uses the calculation principle of INS and the storage function of RNN to estimate the error of INS,thereby obtaining a continuous,reliable,and high-precision navigation solution.The INS/GNSS simulation environment is used to verify the performance of the method.The results show that RNN has a very promising application in the field of combined INS/GNSS navigation and positioning when GNSS signals are blocked.Thirdly,to address the long-term dependency problem that is difficult to solve with current RNNs,and the possibility of gradient explosion or gradient disappearance,LSTM is proposed to be applied to INS/GNSS integrated navigation,and the ADAM optimization method is used to update network parameters instead of traditional non-optimized parameter updates.Methods and simulation results show that the data fusion algorithm can significantly reduce the navigation error during the satellite signal interruption,and the training and modeling are more efficient.Finally,considering that the simulation verification environment is too ideal,the proposed algorithm is verified by on-board test data.The results show that both RNN and LTSM can effectively improve the integrated navigation positioning accuracy when the GNSS signal is blocked under realistic conditions.