Research On MEMS IMU/GPS Integrated Navigation Filtering Algorithm

Integrated navigation systems consist of Strapdown Inertial Navigation System (SINS)based on MEMS and Global Positioning System (GPS) have been the research and development hotspot in the global navigation area because of low-cost, excellent navigation ability and strong complementation, which is of fully wide application prospect. And information fusion is one main direction of SINS/GPS integrated navigation systems’ research and application, it is of great significance to study how to promote the MEMS/GPS integrated navigation system performance by accuracy, reliability and continuity. The contents are as follows:Firstly, the research status on integrated navigation systems and data fusion algorithms is introduced, as well as the fundamental principles of inertial navigation system and INS/GPS integrated navigation system. A third-order damping circuit is adopted aiming at solving the instability problem of INS altitude passageway, and the INS mechanical arrange is provided.The MEMS INS linear and non-linear error models are derivated in detail.Then, the inertial sensors’ error charachteristics are analyzed, including systemic errors and random errors, and random errors of MEMS gyroscopes and accelerometers are analyzed using Allan Variance method. Estimation and compensation of MEMS IMU deviation is realized by using Kalman Filtering, which effectively promote the navigation accuracy and reliability of INS.Thirdly, three filtering methods are utilized on MEMS/GPS loosely integrated navigation system to simulate with practical vehicular experimental data and give the contrast analysis,which include the classical KF, the linear IAE-AKF that can estimate nosie variance through innovation and the non-linear EKF. Results show that the three filtering methods have similar effects, of which KF is proved to be the most optimal, and for the non-linear model under the case of big misalignment, they are all proved to be with the fastest convergence rate and realize the same accuracy and stability as the circumstance of small misalignment.Finally, the random forests algorithm is utilized for data fusion aiming at solving the problem that the MEMS/GPS integrated system accuracy and performance will decrease when GPS signal is lost. During GPS period of validity, Random forest who can reflect the relationship between the input and output variables is build by training or learning. Once GPS signal is interrupted and out, the trained random forest could be used to predict the navigation parameters. Principles and construction methods of decision tree and random forest are researched, as well as how to estimate the forest predicton error by using outbag data. Then how to apply random forest regression into MEMS/GPS integrated navigation system is discussed in detail. Simulation results show that random forest algorithm is of better ability of learning and prediction, and is capable of continuous, reliable and high-precision position and velocity.