This Ph.D. dissertation reports on development of inertial sensors and inertial navigation algorithms for pedestrian inertial navigation applications. Toward enabling such applications, this thesis developed new techniques leading to improvements in gyroscopes and navigation algorithms. The main contributions of this thesis include:
1. Improved structural symmetry and quality factor of Fused Quartz micro-wineglass resonators. We developed an analytical model to predict the resonant frequency of wineglass modes of the resonators with error less than 20%. A model to predict the frequency split of the device was also developed. Directional lapping method was introduced to reduce the as-fabricated frequency split by more than 6X without damaging the integrity of the structure, thus preserving the quality factor of the device. Effects of surface loss were observed and analyzed, and surface reflow was demonstrated to be able to improve the quality factor of the devices. A piezoelectric actuation architecture was explored on the Fused Quartz dual shell resonators, in order to eliminate the metal deposition on the vibratory element to minimize the surface-related energy losses.
2. Identified and quantified, for the first time, energy dissipation mechanisms in Micro-Electro-Mechanical System (MEMS) resonators. By controlling the temperature, air pressure, and surface moisture of the device, viscous air damping, Thermo-Elastic Damping (TED), the anchor loss, and the surface loss of the resonator were manipulated and identified. At room temperature, the quality factors related to viscous air damping, TED, the anchor loss, and the surface loss were experimentally measured to be 625,000, 200,000, 1,350,000, and 1,146,000, respectively. This study provides insight to understanding the dominant mechanism that limits the quality factors of MEMS resonators.
3. Developed an analytical model to predict the effect of IMU noise on the navigation solution uncertainty in the Zero-Velocity-Update (ZUPT)-aided pedestrian inertial navigation. A bio-mechanical model for human walking was built in order to numerically simulate the process. A discrepancy of less than 10% was shown between the analytical and numerical results. Experiments have been conducted and the results were on the same level of the analytical prediction. Among many IMU noise contributions, the dominant factor affecting the accuracy of the ZUPT-aided pedestrian inertial navigation was Rate Random Walk (RRW) of the z-axis gyroscope. This result is envisioned to aid in analysis of the effect of errors in sensors, which might lead to a well informed selection of sensors for the task of ZUPT-aided pedestrian inertial navigation.
4. Improved implementation for the ZUPT-aided pedestrian inertial navigation algorithm for better navigation accuracy. Adaptive threshold based on shock level was first developed for stance phase detection to improve the adaptivity and reliability of the algorithm, demonstrating more than 12X improvement of navigation accuracy. Then, more than 10X reduction in systematic error of position estimation was demonstrated with a careful characterization of the motion of the foot during the stance phases in walking cycles and calibration of IMU in terms of the gyroscope g-sensitivity. Stochastic errors were reduced by 45% when mounting the IMU on the forefoot instead of on the heel.
This work offers new methods and techniques for further reduction of navigation errors by improving the performance of inertial sensors and inertial navigation algorithms.