Research On INS/GNSS Integrated Navigation Algorithm Based On Neural Network

The integrated navigation system composed of Inertial Navigation System(INS)and global navigation satellite systems(GNSS)is widely used in the field of marine integrated navigation.How to improve its accuracy has always been the research focus of scholars at home and abroad.INS navigation is a self-consistent system,which is not easy to be disturbed by external environmental information.Furthermore,INS has low short-term noise and can work continuously.However,due to the solution characteristics and inertial sensor error sources,the navigation nonlinear error will accumulate greatly with time.GNSS navigation can provide high-precision position output for a long time,but its anti-interference ability is poor,resulting in low data update rate.INS / GNSS integrated navigation has the advantages of high precision,low cost and all-weather operation.However,in the case of bad weather,the solution accuracy of GNSS receiver will decline sharply due to electromagnetic interference and shielding of receiving antenna,and the filtering divergence will seriously deteriorate the navigation performance in pure inertial mode.The introduction of neural network assisted filtering method is an important method to solve this problem.After the application of neural network assisted filtering,the navigation can still achieve a seamless solution with high robustness when the GNSS signal is completely unlocked.This paper mainly studies the neural network auxiliary filtering algorithm in INS / GNSS integrated navigation.In view of the fact that the existing neural network auxiliary filtering has no significant effect on the improvement of solution accuracy and reliability and is difficult to meet the requirements of navigation and positioning when the integrated navigation is out of lock,the RBF neural network algorithm improved by RBF neural network and particle swarm optimization algorithm are used for auxiliary filtering respectively.(1)The basic principles of INS and GNSS navigation and the structure of INS / GNSS integrated navigation system are introduced.The attitude solution is compensated by using the accelerometer through the Mahony complementary filter,and the corresponding navigation error solution model is constructed based on the navigation solution characteristics in the geographic coordinate system.(2)The basic theory of standard Kalman filter and extended Kalman filter under its framework is studied.The error solution model of integrated navigation is constructed according to the characteristics of the algorithm.The two-dimensional simulation motion is used to verify the advantages of extended Kalman filter over standard Kalman filter in INS /GNSS integrated navigation.It is also further verified that the accuracy of integrated navigation decreases after GNSS navigation is unlocked,It is necessary to introduce auxiliary methods to deal with it.(3)An integrated navigation algorithm based on feedforward neural network is introduced,which can compensate the output of inertial navigation by using high-precision navigation output.This paper introduces the training process of BP(back propagation)neural network,analyses the shortcomings of BP neural network,and introduces RBF(radial basis function)neural network which can be approximated locally and is also a feedforward network.The basic principle and training method of RBFNN are studied in detail.Aiming at the current situation that the accuracy of RBFNN is not high due to the network parameter learning method,a particle swarm optimization RBFNN(PSO-RBF)algorithm is proposed,and the application of three neural networks in integrated navigation algorithm is compared.(4)The GNSS receiver is out of lock under extreme conditions,which makes the navigation performance decline sharply,so the neural network assisted filtering method is introduced to study.This paper analyzes the advantages and disadvantages of the existing ship navigation neural network auxiliary algorithm.In view of the fact that the effect of BP neural network auxiliary filter on the improvement of solution accuracy in integrated navigation is not significant,RBF neural network and RBF neural network algorithm improved by particle swarm optimization algorithm are used respectively.Through the three-dimensional ship motion experiment,it is verified that the neural network auxiliary filtering technology is effective to solve the problem of rapid decline of navigation accuracy after GNSS signal interruption.It is explained that the RBF neural network auxiliary filtering of particle swarm optimization algorithm has a good improvement effect on the accuracy and stability of the integrated navigation system,and the solution accuracy is quantified and compared.