Investigation Into INS Technique In Frequency Domain

Inertia Navigation System has many unique characters which other navigation systems (satellite navigation, radio navigation, astronomy navigation etc.) cannot provide such as independence, concealment, and providing self-contained motion dynamic at a very high data rate. Although other systems' certain character maybe have much better performance than INS, INS will still be the kernel navigation equipment.Inertia Navigation System (strapdown) accuracy performance mainly depend on inertia sensors and navigation algorithms. In this dissertation, many methods will be introduced to enhance INS accuracy in this two aspects.To improve the performance of lower accuracy gyros mainly depend on modeling gyro drift and then suppress gyro drift by digital filtering, except for improving gyro design and manufacture techniques. In this dissertation, based on analyzing existing digital filters, improved wavelet threshold de-noise method is proposed and used to gyro signal de-noise. Simulate results show that this method has better de-noise effect. For different motion dynamic and environment information, investigate and summarize the principle of choosing wavelet base function and decompose layer in wavelet de-noise.As for strapdown navigation, conventional INS algorithms characters and corresponding disadvantages are analyzed. Modernization method is proposed that is block processing development of the INS algorithms in frequency domain.Using analytical integral substitute for numerical integral because using discrete numerical integral cannot compensate non-commutability errors of finite rotations fully. It needs to reconstruct continuous signal from discrete signal. The source of reconstruction error is analyzed and the error bound is estimated. The method to reduce error is proposes subsequently. Due to sequential data processing in time-domain, consistent with the capabilities of early computers that leads poor efficiency in using current DSP, and the unbalance with GPS data processing, data is processing in frequency-domain in substitute for time-domain.In this dissertation, based on the analyzing the disadvantages of conventional INS algorithms, data block-processing technique in frequency domain is introduced to INS algorithms, and frequency domain implementation of INS algorithms is proposed. Simulation results show that it reduces the non-commutability errors of finite rotations and enhances INS short-term accuracy greatly.