Key Technique Of Multi-source Information Fusion In Integrated Navigation System

Integrated navigation system is the information integrated system used to solve theproblem of navigation and positioning, motion control and equipment calibration alignment.It is the inevitable development trend of network navigation system because of its advantagesof high precision, high stability and high degree of automation. In essence, integratednavigation system is the optimization and fusion system of multi-sensor and multi-sourcenavigation information, whose key technique is the processing and fusion of information. Thecore issues of information fusion can be summarized into three categories: data issue, methodissue and model issue. In order to meet the design requirements, to complete the mandate andmission and to achieve the final purpose, information fusion systems must face the data,method and model of the corresponding object.With the technical background of the design of an underwater vehicle integratednavigation system, and concerned with the three issues of multi-source information fusion, theintegrated navigation system information fusion structures is proposed. Sensor time andspatial registration, adaptive federated filter and its optimality analysis, multi-model fusionare researched in the paper, which are the related key technique for integrated navigationsystem.1. To solve the problem that the time and spatial properties of multi-sensor systemnavigation information are inconsistent, time and spatial registration in integrated navigationsystem is represented clearly. And the source of the time errors and spatial errors and theireffects to the integrated navigation system are analyzed. To solve current time registrationmethod only in the specific condition cannot use in integrated navigation system, a timeregistration method based on segmentation overlap is proposed in the paper. The measurementinformation of the different sensors can be unified into the same time by segment processing.The effective sampling rate and the precision can be improved by overlapping thesegmentation areas. To solve the problem that non-interval federated filter algorithm is unableto suit the project situation with great differences between measurement cycles, an improvednon-interval federated filter algorithm is proposed under asynchronous fusion, which canincrease fusion information and expand the applicability and practicality of the algorithm.Unified mathematics description of the coordinate system transformation in navigation isgiven. According to the analysis of lever arm effect in Inertial navigation system (INS) andthe low-pass filter dealing with the acceleration of the lever arm, the lever arm effect in velocity estimation is obtained by reverse derivation of the lever arm acceleration principle.And the compensation method for velocity estimation by simple model of lever arm error isproposed to solve the effect on velocity estimate by lever arm between multi-sensors.2. Different methods of establishing system equations of the federated filter aresummarized, which is the basic fusion estimation method in the integrated navigation systemin this thesis. The system equations of the INS/ESGM/GPS/LNC/TAN/DVL integratednavigation system are established, and their rationality is predicted by the observabilityanalysis method based on singular value decomposition. By summarizing and analyzingexisted contradictory information distribution principles in adaptive federated filters, and bestudying the relationship between information distribution coefficient and systemperformances, three theorems of information distribution principle are proposed by theoreticalanalysis based on the basic theory of federated filter, which can be used as theory evidence forfederated filter design. From the angle of information source, the problem of informationdistribution coefficient is researched, and a new self-adaptive federated filter algorithm basedon time series analysis is proposed in the paper. The history data of navigation sensors areconsidered and analyzed by autoregressive moving-average (ARMA) model, thus theinformation distribution coefficient is adjusted according to the working status andcircumstances. Simulation results show that higher information efficiency, higher precision,fault-tolerance and stability can be obtained.3. According to the essence of estimate and the intrinsic property of estimate optimality,the optimality of information fusion estimation methods in integrated navigation system isanalyzed. The optimality of Kalman filter in the case of colored noise, Kalman filter in thecase of the correlated noise, decentralized filter in the case of correlated sensor noises,federated filter in the case of correlated estimate, federated filter in the case of differentstructures, and federated filter in the case of different dimensions, are analyzed by theoreticalanalysis and experimental analysis with the optimal rule of linear minimum variance. Thenthe optimality of self-adaptive federated filter based on time series analysis in ideal conditionsand non-ideal conditions are analyzed, which proves that adaptive improvement does notaffect the optimality of the federated filter.4. The method about how to apply the multi-model theory to integrated navigationsystem is researched. The redundant navigation system based on multi-model adaptiveestimation method is researched in the paper, and two ways of analytical redundancynavigation system are analyzed. The multi-model adaptive estimation fusion algorithm inDR/INS redundant integrated navigation system is proposed to guarantee the long time navigation of underwater vehicle. To solve the fusion estimation inaccuracy and long-timefilter divergence caused by model errors, an improved federated filter based on interactivemulti-model is proposed, in which federated filter structure is used to eliminate the correlatedestimate, and interactive multi-model is used instead of Kalman filter. The model error isrestrained by covering the range of the process noise covariance matrix and measurementnoise covariance matrix. And the anti-interference and stability can be obtained when theunderwater vehicle sails in long time.