Research On CNN Fusion Processing Technology For Beidou GNSS/INS Special Area Navigation Application

With the rapid development of science and technology,navigation and positioning technology has been widely used in production and life,and users have put forward higher and higher requirements for the positioning accuracy and stability of navigation and positioning system.Global Navigation Satellite Systems(GNSS)can achieve real-time and all-weather positioning and navigation services,but the positioning accuracy is subject to great external interference,such as in urban canyons,Bridges and tunnels,dense forests and other different environments.GNSS signals are greatly weakened,leading to the decline of positioning accuracy,and even the GNSS signals are out of lock,which leads to the system’s inability to complete positioning and the rapid divergence of navigation and positioning errors in special areas.Inertial Navigation System(INS)can realize navigation and positioning independently and output navigation information independently,but positioning errors increase with the observation time.A single navigation and positioning system can no longer meet the requirements of users on positioning accuracy,availability,stability and low cost under different environments.Therefore,aiming at the requirements of low cost,high precision,high reliability and miniaturization of vehicle navigation System under different environments,this paper develops a low-cost Micro Electro Mechanical System Inertial Measurement Unit(MEMS IMU)and GNSS combined positioning of vehicle navigation system model research,in order to improve the navigation and positioning performance under different environments.The main research contents and achievements of this paper are as follows:(1)Two main filtering models of integrated navigation and positioning under different environments are studied and analyzed:multi-state constrained filtering and nonholonomic constrained filtering.It is found that the multi-state constrained filter model can achieve high precision positioning and navigation by introducing redundant sensors to obtain redundant information,and significantly improve the navigation and positioning accuracy under the premise of increasing the calculation and design difficulty.Nonholonomic constraint filtering uses the carrier motion model to establish the dynamic constraint,which is simple to construct and requires little computation.Due to the uncertainty of the carrier motion,the carrier constraint model cannot be established accurately,and the model is affected by the constraint inaccuracy and other factors.(2)Aiming at the problem that the nonholonomic constraint model does not accurately constrain the lateral and normal velocity of the carrier,a new method is proposed to solve the problem of the accuracy of the nonholonomic constraint model when measuring the lateral and normal velocity of the carrier.In this method,the model is dynamically adjusted to simulate Gaussian white noise with zero mean,and the convolutional neural network is used to optimize the covariance of the pseudo-measurement noise with nonholonomic constraints,so as to achieve the best fusion of Kalman Filter model.(3)Based on the nonholonomic constraint method and the convolutional neural network prediction model,the data model processing architecture of this paper is constructed.In view of the divergence of navigation and positioning errors in different areas and the requirements of the project research and development plan,the design and research of the INS/GNSS integrated navigation system model for low-cost multi-system INS and GNSS was carried out.The algorithm model is verified and evaluated through the measured data.The experimental results show that the convolutional neural network as the auxiliary prediction technology in this paper can effectively solve the problem of inaccurate speed constraints in the nonholonomic constraint algorithm,aiming at the problem of inaccurate speed constraints.Thus,the positioning accuracy and stability of integrated navigation system in complex environment are greatly improved,which has important practical application value.