Research On Information Fusion And Fault Tolerant Optimized Method Of Surface Ship Navigation System

The navigation system is an important guarantee for surface vessels to complete various types of operations safely,economically,and autonomously in a variety of complex navigation environments.It has been discovered that the Micro Inertial Measurement Unit(MIMU)navigation sensor has been widely used in the navigation field of aviation,aerospace,and other vehicles.However,there are some limitations in its application in surface vessel integrated navigation technology.The addition of MIMU modules to maritime autonomous surface ships to achieve robust and fault-tolerant navigation and cooperate with other commonly used navigation equipment to achieve robust ship motion status output when shortterm failures occur.Following issues must be addressed:(i)In order to address the issue that the low measurement accuracy of MIMU can exacerbate the nonlinearity of the surface vessel’s integrated navigation system,an appropriate integrated navigation nonlinear filtering algorithm must be chosen;(ii)It is necessary to design a fault detection and fault tolerant optimization algorithm for the surface vessel’s integrated navigation system in order to not only realize robust integrated navigation with multi-source sensors,but also to detect faults in the sub systems in real time and isolate these faults when necessary;(iii)If the global navigation satellite receiver is unable to provide accurate motion status signals,it is necessary to ensure the accuracy of the estimation of vessel motion state by the sub filters to assist the sailor in better understanding the vessel navigation situation.Based on the above analysis,the research work in this dissertation primarily includes the following main contents:(1)To carry out the simulation research on the navigation sensor system of the surface vessels,the kinematics and dynamic model of the surface vessel is studied,based on which,the navigation measurement error model of magnetic compass,inertial navigation system,global navigation satellite system,and Doppler velocity log,are established.The error between the higher order terms of the Taylor expansion and the true value is analyzed,Extended Kalman Filter(EKF),Unscented Kalman Filter(UKF),and Cubature Kalman Filter(CKF)under the different state estimation dimensions in the process of surface vessel integrated navigation filtering.Following the theoretical and simulation verification,CKF is chosen as a sub filter with perfect theory,good accuracy,and robustness.(2)Given the problem of the CKF sub filter’s dependence on zero-mean Gaussian white noise,the simplified sequential Sage-Husa algorithm is used to optimize the CKF in order to improve state estimation accuracy.Given the shortcomings of the simplified sequential SageHusa CKF algorithm,such as the dependence of measurement noise covariance threshold setting on experience and obvious estimation error chattering,the H∞ algorithm has been applied to the sequential Sage-Husa CKF,and an adaptive operator is proposed to improve the algorithm’s robustness to unknown noise.According to the simulation results,the robust adaptive filtering algorithm proposed in this dissertation not only has a slightly higher state estimation accuracy than the sequential Sage-Husa CKF,the robustness of the system is enhanced while avoiding estimation error chattering.(3)Given the requirement for high fault tolerance in the algorithm for the integrated navigation system of surface vessels,the information weight of each sensor must be allocated fairly.As a result,an observability algorithm of improved linearized integrated navigation based on CKF has been proposed,which not only considers the system’s observability,but also the accuracy of external measurement information.After analyzing the Strapdown Inertial Navigation System/Global Navigation Satellite System(SINS/GNSS)and Strapdown Inertial Navigation System/Doppler Velocity Log/Compass(SINS/DVL/COMPASS),the results show that the full state of the SINS/GNSS sub system is observable and the heading observability is the weakest.The longitude error is not directly observable in the SINS/DVL/COMPASS subsystem,and the latitude error observability is low.Finally,a robust adaptive federal filtering algorithm for surface vessel integrated navigation based on the observability of the integrated navigation system has been proposed,with SINS/GNSS position error incorporated as a non-public state.Real-world ship data show that the proposed algorithm is more robust than the ARCKF used in the sub system.(4)Given the fact that faults in the sub system can pollute the output of federal filter and impair navigation performance.Under the CKF framework,an H-/H∞ fault detection algorithm has been proposed that takes into account both the robustness of residual to noise and the sensitivity to faults.Given the shortcomings of insensitivity to gradual faults and reliance on a priori knowledge of the residual observer,a cascaded fault detection network comprised of long-and short-term work(LSTM),multi time-domain feature fusion fault detection network,and Short Time Fourier Transform(STFT)frequency domain fault detection network has been proposed.It is discovered that the constructed cascaded network is not only sensitive to soft and abrupt faults,but also does not rely on prior knowledge to set the threshold and sliding window width,and the detection result is unaffected by the value of the extreme residual evaluation function.(5)Aiming at the fusion of real-time observation information of marine radar and hydrological static information of electronic chart,a Radar/Electronic Chart perception information fusion algorithm based on the deep learning algorithm has been proposed.To begin,a deep learning algorithm is used to detect the region of interest in radar images of feature targets.Then,based on the detected radar feature target region of interest and the radar operating principle,a feature point search algorithm has been proposed to perform fusion based on the searched mapping feature points.Then an algorithm based on weighted Harr discrete wavelet transform is proposed to optimize the fusion result,which enhances the highfrequency information of electronic charts and retains the characteristics of real-time water surface information detected by radar.It has been discovered,using real ship radar images and chart data,that the algorithm can realize feature-level fusion of radar image and electronic chart,allowing sailors to better understand the navigation situation.(6)Given the low accuracy of SINS/DVL/COMPASS position error estimation,a characteristic inflection points detection algorithm based on double curvature and distance constraints has been designed to detect several characteristic inflection points in electronic charts in order to estimate the change of vessel position coordinates.As an external observation,the position change is fed into the SINS/DVL/COMPASS.Finally,the simulation is performed using the SINS/DVL/COMPASS/VISUAL information fusion algorithm proposed in this dissertation.According to the simulation results,when there are abnormal values in DVL,the SINS/DVL/COMPASS/VISUAL integrated navigation algorithm is more robust.