Reseach Of GPS/DR Vehicle-mounted Integrated Navigation System Based On Multi-sensor Information Fusion

Multi-sensor information fusion (MSIF) is the basic theory of information processingin the integrated navigation system. It is an effective solution to achieve high precisionpositioning performance of integrated navigation that through the perfect vehicleintegrated navigation system structure of information fusion and informationprocessing algorithms. According to the needs of vehicle navigation system precisionand continuity, as well as the constraints of volume, power consumption and cost, theGPS/DR vehicle integrated navigation system become the research hotspot in thevehicle navigation.Kalman filtering is the most commonly used method in processing navigationestimation. Among them, the extended Kalman filtering (EKF) and unscented Kalmanfilter (UKF) to solve the filtering problem of nonlinear systems. However, they do nothave adaptability, when work in complex filtering environment. However, they haveweak adaptability when work in complex filtering environment. This paper uses theimproved adaptive algorithm to solve this problem, and has mainly done the followingworks:Firstly, this paper expounded the multi-sensor data fusion theory, GPS and DRpositioning principle of the system and analysis of integrated navigation system error.Studied the structure and process of basic Kalman filter (KF), extended Kalman filter(EKF), unscented Kalman filter (UKF) and recursive least squares (RLS) adaptivealgorithm.Secondly, this paper analyzed the problem of EKF and UKF algorithm prone tolocate lost reliability, or even filtering divergence, which is caused by the changes sizeof the environmental noise. In order to solve this problem， the EKF and UKFalgorithm is improved, and get the adaptive extended Kalman filter (AEKF) andadaptive unscented Kalman filtering (AUKF) algorithm.Thirdly, according to the tight combination of GPS/DR integrated navigationsystem, at the basis of the vehicle’s dynamic characteristics and the acceleration of current statistical model, and then establish system mode suit Kalman filter processingwhen vehicle works in the earth in a two-dimensional rectangular coordinate.Finally, using the MATLAB software simulation experiment, and from twoaspects of convergence speed and accuracy proved the validity of the AEKF andAUKF algorithms.