A Study On Fusion Filter Algorithm For Low Cost Tightly-coupled AHRS/GPS

With the development of MEMS technology, the micro inertial measurement uint (MIMU) is widely used in solid-state low cost strapdown attitude and heading reference system (AHRS). AHRS is a three dimensional measurement package including MIMUs and magnetic sensors, which presents the prominent characteristics of small volume, light weight, high reliability and strong shock-resistantance, accomplishing the real-time attitude and heading control for carriers under dynamic conditions. Moreover, the ideal positioning mode of integrated navigation is achieved by combine AHRS with a miniature GPS.The extended Kalman filter (EKF) is the effective means when solving data fusion issues of integrated navigation system. Due to the large drifts and poor stability of MEMS sensors, the fusion filter solution is proposed, which fuses the improved EKF methods and intelligent control ones based on artificial netwoks, achieving the optimal estimates of parameters. The dissertation focuses on two parts, namely operation technique on stand-alone AHRS and fusion filter research on tightly-coupled AHRS/GPS.The strapdown attitude algorithm of micromachined AHRS is researched in-depth based on mastering the operation principle of these types of sensors. Two non-colinear referece vectors, which involve gravity from accelerometers' measurements and geomagnetic field from magnetic fluxgate sensors' measurements, are observed to correct the attitude and heading obtained from gyros' integral. In sequence, the observation update equation of Kalman fiter is constructed in body frame to estimate and compensate gyros' zero bias in real time.Considering that the attitude updating might be over rapid for land vehicular carriers, shortening attitude updating period is of great essentialness. The dual quaternion (DQ) algebra is introduced, optimizing the calculating process for extracting attitude updating matrix. This leads to calculate the rotation matrix with higher rapid, and provide the estimates of the translation at the same time. The approach is verified to be feasible by simulations under the straight and swing motion base of MEMS AHRS.Since AHRS relies greatly on the geomagnetic field, the magnetic disturbances and corresponding solutions are discussed. In addition, the software scheme on magnetic heading error compensation is analyzed. It is proved that the method is an easy one, and it adapts to the practical applications.The adaptive quadratic EKF (QEKF) algorithm based on strong tracking (ST) theory for tightly-coupled AHRS/GPS is proposed, and the parameter model is established by fusing pseudo_range-pseudo_range rate-heading measurements. The quadratic truncation error extracted from linearization is compensated by using QEKF, approaching the nonlinear feature in nature. Through ignoring the prevenient states information based on strong tracking algorithm, the sudden changes of states are rapidly been sensed so as to guarantee the astringency, enhancing the robustness of the whole system.The neural network aided filter is designed in virtue of the improved EKF model, indicating the fusion feature of intelligent control methods and classical ones. Considering the large training sample and no-feedback net structure, the generalized radial basis function neural network (RBFNN) is selected to modify the predicted states on line. In consequence, the accuracy and kinematic performance are all enhanced as expected. Study has shown that the fusion filtering is feasible for parameter estimates in the field of low cost AHRS/GPS integration, even if the accuracy of the sensors is modest.