| Due to the shortage of mineral resources and the development of science and technology,the polar region has become a strategic priority area for resource exploitation and military expansion.The grid inertial navigation system(INS)can surmount the lack of northward direction,error amplification,and calculation overflow,which becomes the pre-requisite of polar arrival and voyage,and the warranty of essential security for the various carriers.Moreover,the polar transfer alignment is critical to ensure the precision and speed of attitude initialization of grid INS under the moving base.Since the Arctic passage has developed till now,the strategic focus gradually has shifted from maritime security to combat applications.Recently,the grid transfer alignment cannot self-adapt to different carrier platforms of alignment and dynamic environments,restraining the performance and application scenario.The polar geophysical field impairs the optics-based and electromagnetism-based navigation systems,and the master INS calibrated via these systems may happen to errors,causing the low and unreliable quality of observations of the polar transfer alignment.The complicated and changeable marine environment easily results in inappropriate statistical characteristics of observations and inaccurate mathematical modeling,which directly degrades the filter estimation,amplifies the estimated deviation,and weakens the fault-tolerant ability and environmental adaptability.The dissertation explores improved grid transfer alignment,device-assisted polar transfer alignment,and signal estimation,starting with modeling and analysis.The target is to adapt to different carrier platforms and balance precision and speed.The thesis involves research contents as follows.Initially,the thesis combs the critical problems and designs its solutions through the applicability analysis of the nonlinear method of grid transfer alignment in terms of the mathematical model consistency,nonlinear system observability,and error characteristic.Based on the description of the grid INS mechanism,the grid transfer alignment is modeled under the large azimuth,which decomposes the attitude matrice between the master and slave body frame to seek these intrinsic connections and picks out a better arrangement for the polar region.Beginning with global observability,whether observing the grid transfer alignment is theoretically argued to search for a superior maneuver.The thesis digs fundamental problems and conceives solutions via deriving the effect of strong nonlinearity on the error model and filter estimation.Secondly,targeting the input deviation of the mathematical model and the biased estimate of filter caused by forced linearization of horizontal misalignments,the dissertation proposes an improved grid transfer alignment via constructing nonlinear error equations based on the multiplicative error quaternion and optimizing its nonlinear filter.The multiplicative error quaternion redefines the misalignments to derive the systematic state equations and construct the "grid velocity+error quaternion" observation equations,eliminating the assumption of small misalignments from the mathematical model and equipping adaptability to different carrier woking conditions.For non-additive noise,normalization,and negative-define error matrix caused by multiplicative error quaternion,the thesis modifies the augmented unscented Kalman filter(UKF)optimized algorithm via establishing the attitude cost function and depression of order of multiplicative error quaternion.The simulation and semi-physical experiments are designed to validate the proposed algorithm with high-precision and solid self-adaptability.Subsequently,for uncontrolled observation quality caused by polar unique geophysical field,an innovative bionic polarized-light navigation technique is explored,and an improved polar transfer alignment assisted by bionic polarized-light navigation is proposed.Since the azimuth solved by polarized-light navigation amplifies with the increasing latitudes,this work raises an improved polarized-light navigation algorithm modeled under the grid frame and designs a weighed clustering to upgrade the precision and unify the coordinate datum of various navigation devices.Through extracting the zenith point from the skylight polarization pattern to offset the horizontal attitude error of polarized-light navigation,an improved transfer alignment assisted by bionic polarized-light navigation in the polar region is put forward via constructing the "grid velocity+polarized-light attitude" observation model.The simulation and semi-physical experiments are designed to validate the proposed algorithm.Eventually,a proposed signal estimation for grid transfer alignment based on the BP neural network(BPNN)solves the estimation deviation of the nonlinear filter caused by the inappropriate extraction of signal statistical characteristics and inaccurate mathematical model in the complicated polar marine environment.The question of the effect of noise on signal estimation is posed through theoretical derivation and simulation analysis.The BPNN is designed by dissecting the working principle of the polar transfer alignment from the aspect of the learning method,learning algorithm,and network structure.Moreover,for the poor generalization,the training samples are optimized to match the improved algorithm of gird transfer alignment,and the pre-processing of data is designed to eliminate the uneven data distribution caused by different dimensions.The augmented UKF assisted by BPNN is raised to offset the relative deviation.The semi-physical experiments are designed to validate the prediction performance of the designed BPNN and the proposed signal estimation. |
PDF Full Text Request