Federated Filter Base Information Fusion Method For Position Reference System

Dynamic Positioning (DP) Ship mostly engaged in high risk marine engineering work.The international Maritime Organization and every National Register of Shipping demandthat DP ship should be equipped with multiple position reference system (PRS). Therefore, itis a problem needs to be solved that how to effectively use a variety of measurementinformation to improve the security and operating performance of the DP system. Informationfusion technology is a kind of effective method which could optimize the information frommulti-sensor. In this paper, in order to deal with the DP vessel PRSs, a federated filter whichbased on the unscented Kalman filter (UKF) method is used to achieve the multi-sensorinformation fusion function. This method could overcome the weakness of single PRS andmake use of the data from multi PRS of DP ship effectively through information fusion.Meanwhile, by using the UKF technique, the error in the nonlinear estimation problem whichinduces by the linearization process of traditional extended Kalman filter (EKF) is also solved.Additionally, the federated filter method used in this paper can not only achieved theinformation fusion function for the DP vessel PRSs but also significantly enhanced the wholePRSs’ fault tolerance and fault recovery capability.The DP ship multi redundant PRS is a kind of typical multi-sensor system, while themeasurements of diversified PRSs exists highly nonlinear characteristic. Most of thetraditional multi-sensor data fusion method adopted the noise covariance to describe thecharacteristic of the sensor and applied it in the research of fusion algorithm. Different fromthe traditional method, the nonlinear measurement equation which derives from the measureprinciple of each kind of PRS is used to describe the sensor characteristic in this paper.Additionally, the federated structure and UKF method are used to accomplish themulti-sensor data fusion function for the DP vessel PRSs.The federated filter contains one master filter and several local filters. It is a specialdistributed filtering method which contains two data processing steps. Aiming at the DPvessel PRS (include taut wire, acoustic and GPS), the local filters are designed by using theUKF method which employed the nonlinear measurement equation derived for each PRS andthe nonlinear ship motion equation. The master filter is designed based on the informationthat contains each local filter’s covariance matrix of estimation error and estimated states toachieve information fusion. In order to improve the whole method’s ability to deal with theerror in the PRS and have a certain extent of malfunction resumes function, a fault detection module is introduced before the master filter in the federated filter structure that with nofeedback. By using this method, a fault-tolerant federated filter is designed which with faultdetection and recovery capabilities.Finally, the established simulation environment is used to verify the function of eachmodule in the design of two kind of federated filters which uses the structures with or withoutfeedback. The simulation results show that the data processing and information fusionfunction for the PRS can be achieved excellently by using the UKF federated filter method.Also, by introducing the fault detection module, the overall information fusion method’s faulttolerance ability is improved. Therefore, the proposed UKF fault-tolerant federated filtermethod can achieve the information fusion function for the DP vessel PRSs, it provides aneffective approach to improve the DP vessel PRSs’ security, dependability and precision.