An Intelligent Calibration of SINS Using Neural Networks on Moving Base

Abstract

In order to weaken the error of inertial sensors and to improve assaulting precision of an air launched missile, the technology of neural networks was attempted to on-line calibration of Strapdown Inertial Navigation System (SINS). Aiming at the time-varied specialty of SINS on moving base, an input-output sample structure was proposed to treat the neural networks for calibrating and revising the error of inertial instrument. Consequently, when a missile was appending under the wing, the trained neural networks can be straightway used for automatic calibration in the free-flight phase; In order to resolve inconsistent measurement of gyroscopes and accelerometers when a missile was appending under the wing and in free-flight phase modes, the error angles between master and slave SINS were estimated in advance, then the input sample of neural networks can simulate the free-flight phase. As a result, the precision of inertial sensors can be greatly improved, and the simulation results indicate that the intelligent calibration method is feasible.