Research On Shipborne Integrated Navigation Algorithm Based On Neural Network

With the rapid development of unmanned ship industry and integrated navigation technology,integrated navigation system based on inertial navigation system/global navigation satellite system(INS/GNSS)have been widely used for its convenience in engineering practice.However,when GNSS is denied,the INS/GNSS integrated navigation system will be converted to a pure inertial navigation system,resulting in a sharp decline in navigation accuracy.To solve the problem,the thesis studies the integrated navigation algorithm in GNSS denied environments based on the INS/GNSS integrated navigation theory framework model.In order to improve the positioning accuracy of integrated navigation system in GNSS denied environment,the thesis first proposes the sample features extraction algorithm,which is used for the inertial navigation calculation and filtering fusion of received sensor data to obtain the feature samples required by model training and prediction.On this basis,the error correction algorithm of inertial navigation system is further proposed,which is responsible for error modeling or error calculation of the sample data output by the sample features extraction to correct the inertial navigation system output data in GNSS denied environments.On the basis of studying the motion characteristics of inertial navigation system,aiming at the influence of random noise on the attitude characteristics in the moving state,the attitude feature extraction algorithm based on time-varying complementary filtering is proposed.The similarity between the random noise in the moving state and the root mean square value in the stationary state is used to assume the random noise,which effectively solves the problem of drift in the solution result under the fixed coefficient.Aiming at the problem that the estimation error of the kalman filter increases and the filtering result diverges when the output measurement noise is constant,an adaptive algorithm based on new information is used to judge the uncertainty of the process model and an improved state prediction covariance matrix is constructed.When the process model is uncertain,the state prediction covariance matrix is adjusted in real time to realize the accurate solution of the target motion state.The error correction algorithm of inertial navigation system is further proposed in GNSS denied environments.When GNSS signal is available,the error model of inertial navigation system is established.When GNSS signal is interrupted,the saved error model is used to predict the system error and feed it back to the inertial navigation system for error correction,which improves the positioning accuracy of the integrated navigation system in GNSS denied environments.Based on the theoretical framework model of the INS/GNSS integrated navigation system and the above theoretical research,the thesis designs the overall architecture and each module of the INS/GNSS integrated navigation system in detail according to the actual needs,applies the integrated navigation algorithm to the system and verifies the algorithm with the actual data.The research results show that when GNSS signal is denied for 30 s,the east position error of the proposed algorithm is reduced by 85.6% compared with the pure inertial navigation system,and the north position error is reduced by 84.9% compared with the pure inertial navigation system.Therefore,the proposed algorithm can effectively output the integrated navigation data and improve the positioning accuracy of the integrated navigation system in GNSS denied environment.