Abstract: An improved federated Kalman filter configuration was designed based on the analysis of the features of GPS/DR integrated positioning system. A conventional Kalman filter is used to process the information from the linear GPS subsystem, while an unscented Kalman filter is used to process the information from the nonlinear DR subsystem. The main filter makes the optimal estimation of the results from the subsystems using the dynamic informationsharing coefficient, and compensates the error of DR subsystem to reduce the accumulative error of DR. The test results showed that the proposed algorithm meets the positioning request of the onboard navigation system and keeps a good reliability and a high positioning accuracy when the GPS positioning is poor.

Key words: engineering of communication and transportation system, data fusion, integration of GPS/DR, Kalman filter, federated filter, analysis with varied weight