Research On Data Fusion Algorithm Of GPS/DR Integrated Positioning System

As an important aspect of Intelligent Transportation System, vehicle positioning system has great significance in many areas such as alleviating traffic, convenient for driving, transportation management, guard against theft and giving alarm, urgency asking for help. A key point of Vehicle Positioning System is to choose positioning way which aims to gain accurate and reliable information of vehicle location. GPS/DR integrated positioning way not only can solve the problem that solitary GPS is unable to position due to singal being shielded, but can restrain the cumulative error of DR effectively. Accuracy and reliability of positioning system are improved greatly , so this way has been adopted widely. However, for the sake of cost, cheaper DR sensors are adopted usually in GPS/DR integrated positioning system. So fusion algorithm is indispensable to improve the performance of whole system. That is how to fuse location information of GPS and DR effectively. Therefore, the key point of achieving GPS/DR integrated positioning is data fusion scheme, Kalman filtering is a better way.Data fusion algorithm of GPS/DR integrated positioning system is studied detailed based on Kalman filtering theory. Firstly, filtering model of integrated positioning system is founded based on vehicle current statistical model. Appplication of Extended Kalman Filtering(EKF), Federated Kalman Filtering(FKF) and Strong Tracking Filtering(STF) in the system are studied and analyzed respectively.Whereas the base of these algorithms is EKF, which has many drawbacks of slow flitting convergence rate , poor robustness to system model error and noise statistics, uneasy application in practice. In view of this, Unscented Kalman Filtering as a novel nonlinear filtering method is introduced. For it is not necessary to linearize nonlinear system, Jacobi matrix computing and linearization error are avioded, so its practicability and filtering estimation accuracy are better than traditional EKF. Meantime, according to characteristics of GPS/DR integrated positioning system, UKF is simplified(SUKF) for increased computational efficiency. And Federated Unscented Kalman Filtering (FUKF) as well as Strong Tracking Unscented Kalman Filtering(STUKF) are proposed by combining UKF with FKF and STF perfectly. Through detail computer simulation and analysis, it showed that the filtering precision, convergence rate, robustness, practicability and reliability of UKF all improved greatly compared with traditional EKF. Especially, FUKF and STUKF not only keep merits of FKF and STF, but fuse higher accuracy and stronger robustness characteristics of UKF, it meet the requirements of positioning system with low cost and higher accuracy indeed, and possesses certain practical value.