Inertial based Integration with Transformed INS Mechanization in Earth Frame

This paper proposes to use a newly-derived transformed inertial navigation system (INS) mechanization to fuse INS with other complimentary sensors. Through formulating the attitude, velocity and position as one group state of group of double direct spatial isometries, the transformed INS mechanization has proven to be group affine, which opens door to log-linearity and filtering consistency. In order to make use of the transformed INS mechanization in inertial based applications, both the right and left error state models are derived. The INS/GPS and INS/Odometer integration are investigated as two representatives of inertial based applications. Some application aspects of the derived error state models in the two applications are presented, which include how to select the error state model, initialization of the SE2(3) based error state covariance and feedback correction corresponding to the error state definitions. Land vehicle experiments are conducted to evaluate the performance of the derived error state models. It is shown that the most striking superiority of using the derived error state models is their ability to handle the large initial attitude misalignments, which is just the result of log-linearity property of the derived error state models. Therefore, the derived error state models can be used in the so-called attitude alignment for the two applications.