A designed AKF algorithm applied to unconventional GPS and multiple low-cost IMUs integration strategy

This research applied an unconventional KF that directly estimate navigational parameters instead of the error states to integrate GPS receivers and multiple low-cost IMUs. In practice, the a-priori variance matrix of the process noise vector and the measurement vector are unknown or approximated, which may produce unreliable results. With this in mind, this research proposed an adaptive Kalman filter (AKF) algorithm based on variance components estimation, to simultaneously estimate the variance matrix Q and R by taking advantage of the measurement residuals and the process noise residuals and the measurement redundancy contribution. Besides, the weights of measurements from each sensor were calculated by the posterior variances so that the function of each measurement can be reasonably distributed in Kalman filter, achieving a better structure for the fusion algorithm. Moreover, the systematic errors and measurements of these multiple IMUs were individually modeled instead of being a group of the commonly shared states for all of the IMUs. The real-time raw outputs of multiple IMUs and GPS simulated were processed to demonstrate the performance by utilizing the unconventional integration strategy and the designed AKF algorithm based on variance components estimation.