Identification and estimation of stochastic error parameters of mems inertial measurement unit

Inertial sensors commonly used for navigation systems have specific error characteristics, classified as deterministic and stochastic errors. The error parameters directly affect the inertial sensor performance. Specifying the error characteristics of an inertial sensor depends on the quality of identification and estimation processes. In this thesis, inertial sensor parameters were identified depending on the frequency-domain and time-domain approaches. Revealing the stochastic sensor parameters using the frequency-domain approach (i.e. power spectral density-PSD) is more complex and challenging to apply as compared with the time-domain approach (Allan deviation based). Because of this reason, time-domain methods were investigated in more detail, and new approaches were developed to identify stochastic error parameters. Stochastic error parameters were identified by lines with different slopes on the Allan deviation curve in the time-domain based on the mathematical relationship established between the Power Spectral Density function and the Allan deviation. The new methods used in this thesis are built on the least squares algorithm and the least squares algorithm with a forgetting factor. The stochastic error parameters of each inertial measurement unit (IMU) axis were analyzed from the Allan deviation values of the IMU data collected in a certain time period via the least squares algorithm. Thanks to the least squares algorithm with forgetting factor, analysis of the stochastic error parameters was performed recursively and adaptively. Besides, parameters were obtained using a certain amount of current data depending on the value of the forgetting factor. In addition to that, model-based simulation results and real-time results of Kalman Filter algorithm are also shared in this thesis. Keywords: Allan deviation, inertial measurement unit, Kalman filter, least square algorithm, stochastic errors