Observability Analysis of Initial Alignment and Its Accuracy Improvement

Yufeng Zhang,Yang Gao

2005-01-01

Abstract:Traditional initial alignment technique for Inertial Navigation System (INS) often employs position, zero velocity and zero rotation on eastward axis as observables, which however will turn into partially observable once the biases of inertial sensors become random constant. This will result in degradation to the performance of initial alignment. In this paper, a novel alignment technique will be proposed which could make the alignment system completely observable even if the inertial sensor biases become constant. This technique will introduce the constraints of gravity and earth self-rotation as part of the observables in addition to the measurements used by traditional alignment method. The observability of this new alignment system will be investigated theoretically and the accuracy comparison between the traditional and new alignment will also be conducted theoretically. Computer simulations will be conducted to verify the theoretical development and assess its performance. The results indicate that the new alignment system can keep completely observable no matter what kind of the random process is for the sensor biases and the alignment performance can be improved after the application of the gravity and Earth self-rotation constraint.