Performance Improvement of Inertial Navigation System Based on Frequency Domain Approaches

Abstract

In this dissertation, we discuss a performance improvement method in which frequency domain approaches are applied to an inertial navigation system. The conventional inertial navigation system (INS) algorithm computes all navigation solutions in the time domain. The conventional INS has limitations in it is impossible to resolve some problems by computation in the time domain, such as the observability problem of transfer alignment and the coning/sculling error mitigation algorithm. Thus, frequency domain approaches are applied to resolve the problems. For the first topic, the frequency domain approach is applied to the transfer alignment method. Typical transfer alignment algorithms estimate the initial attitude and velocity of a projectile using the INS result of the launcher as a measurement. Typical methods have low accuracy for a steady state launcher due to the lack of observability. Thus, a new transfer alignment algorithm using the frequency domain approach is proposed in order to solve observability problem. Amplitude measurements obtained by using a discrete time Fourier transform are applied to transfer alignment algorithm as new measurements and these serve to augment the conventional transfer alignment filter. In simulational results, yaw estimation performance improved with a vibration signal. The proposed algorithm is superior to conventional algorithms in rapid transfer alignment of a steady state vehicle. For the next topic, a coning/sculling error compensation algorithm is developed. Typical coning and sculling compensation algorithms mitigate the coning and sculling error by using the cross product between inertial sensor measurements. These methods have limitations when applied to vehicles with large coning and sculling errors, because they have theoretical performance limits. Thus, a new direct coning/sculling compensation algorithm is proposed. The continuous coning and sculling error can be expressed by using vibrating parameters such as frequency, amplitude, and phase differences. Our new method uses adaptive notch filtering that estimates the parameters in sinusoidal signals on gyro outputs. If there are continuous sinusoidal signals on gyros, an adaptive notch filter can be used to estimate the parameters of the sinusoidal signals, and a direct coning/sculling compensation algorithm computes the attitude correction term produced by the coning/sculling signal. Our simulation results show that the proposed algorithm works appropriately.