Tightly Coupled GPS/SINS Filtering Algorithm Based On Rao-Blackwell Paticle Filter

Because the Inertial Navigation system(INS) and the Global Navigation SatelliteSystem(GNSS),can improve the accuracy of the navigation system, make up for theshortcomings of using single navigation system, it is the main direction of navigation researchfield nowadays. In the research of the integrated navigation system, the information fusion isalways the key and difficult point of the research. Although the kalman filter had been puttedforward as theoretical support for inertial navigation system and satellite navigation system’scombination, but in the actual instance, the system is always nonlinear system, in order toimprove the accuracy of integrated navigation system, scholars have proposed manyimproved filtering algorithm. Such as Extended Kalman Filter(EKF), Unsented Kalman Filter(UKF), H filter, Particle filter (PF) and the filtering algorithm combination with neuralnetwork, and improve the filtering algorithms in different degree. This paper mainly studyabout GPS/SINS integrated navigation system and the filtering algorithom of it.In this paper, the principle of the GPS navigation system and the Inertial navigationsystem are introduced at first. Then the error of GPS navigation system and Inertialnavigation system are analyzed and the error models are established, the state model andmeasurement model of the GPS/SINS loose coupled integrated navigation system andGPS/SINS tightly coupled navigation system are established based on the contents alreadyintroduced. Secondly the classical filtering algorithoms of the integrated navigation systemare introduced clearly, thereinto the unsented kalman filtering algorithm and the particlefiltering algorithm can work without linearizing the system, but the Unsented kalman filteringalgorithom depend on the gaussian distribution; although the Particle filtering algorithom iscompletely suitable for nonlinear, non-gaussian system, but the particle filtering also hasshortcomings such as particle degradation, particle depleted and the of high amount ofcalculation. The particle filtering algorithom which was putted forward, mainly includedfiltering algorithom based on improving the importance function of the particle filter, thefiltering algorithom based on inducting of resampling algorithom, and intelligent particlefiltering method. This paper use unscented RTS smoothing algorithm as the important densityfunction to improve the accuracy of the integrated navigation system’s information fusionalgorithom, and adopt the Rao-Blackwell theorem to reduce the calculated account of the highdimension integrated navigation system. At last, this paper use MATLAB7.1to simulate andanalyse the Tightly Coupled GPS/SINS navigation system, the results shows the new filtering method has better representation than the UKF algorithom in improving the accuracy of thesystem.