Research On Underwater Vehicle Inertial/Acoustic Information Integrated Navigation Algorithm

The sensors used in navigation and positioning of underwater vehicles usually include IMU,GNSS receiver,and sensors based on underwater acoustic information such as DVL and forward-looking sonar.The underwater vehicle can receive GNSS signal during the in-motion alignment after placement and during the correction of position error after buoyancy,but GNSS positioning information is not available during diving,and the underwater vehicle can only conduct autonomous navigation based on IMU and underwater acoustic information sensors at this time.This paper mainly includes the following research work:(1)Lie group extended Kalman filter(LG-EKF)is introduced.And the initial LG-EKF is improved by replacing the earth-centered earth-fixed frame with the tangent plane frame centered with the starting point and replacing the earth-centered inertial frame with the inertial frame corresponding to the tangent frame,and new differential equations of error state are derived.Based on LG-EKF,the observation equation of INS/DVL assisted with full GNSS for in-motion alignment is established.In view of the situation that GNSS positioning information can only be received at the initial and final moments,the virtual position observation is obtained by interpolating the entire position based on the dead reckoning of the gyroscope and DVL.The in-motion alignment experiments prove the superiority of the LG-EKF algorithm over the traditional EKF algorithm and the feasibility of virtual position observation.(2)Adaptive Kalman filter(AKF)is introduced,which estimates the covariance matrix of observation noise online based on the innovation sequence.The LG-AKF algorithm is designed by combining LG-EKF and AKF.Experiments of INS/DVL combination based on actual measuring data over a long sailing time prove that the LG-AKF algorithm is superior.LG-AKF can deal with angle error better than AKF,especially in the case of larger initial attitude error.(3)The interference problem of DVL to forward-looking sonar images in some special engineering applications is discussed.A solution of turning off DVL,extracting carrier velocity directly from sonar images,and making it serve for integrated navigation is proposed.According to the specific application conditions,the elevation zeroing model and altitude simplifying model are proposed to estimate the carrier velocity from the motion of feature points.A method combining optical flow tracking and outliers elimination is designed to realize data association between Blueview M900 sonar images.In view of the water conveyance tunnel inspection in the South-to-North Water Diversion Project,positioning experiments are carried out to prove the effectiveness of inertial/forward-looking sonar loosely coupled integration based on velocity extraction.(4)The inertial/forward-looking sonar tightly coupled integration algorithm is designed by adapting the multi-state constraint Kalman filter(MSCKF)which has been used successfully for vision-aided inertial navigation to make it suitable for the underwater environment.The state propagation and state augmentation processes of the filter are derived,and the observation equation corresponding to re-projection errors of the feature points in sonar images is established.To solve the problem of global 3D coordinate estimation of feature points in the process of establishing observation equations,the related formulas of linear triangulation and iterative optimization based on the least-squares method are derived.The results of experiments based on actual measuring data show that the positioning accuracy of the inertial/forward-looking sonar tightly coupled integration is also similar to that of the INS/DVL integration.