Accurate Identification and Implementation of Complicated Stochastic Error Models for Low-Cost MEMS Inertial Sensors

To enhance the inertial sensors performance, accurate characterization as well as modeling of the corresponding stochastic errors are required. In addition, accurate position and orientation information can be achieved when integrating inertial sensors with other position fix systems, such as Global Navigation Satellite Systems (GNSS). This paper investigates the Generalized Method of Wavelet Moments (GMWM), as a Wavelet Variance (WV)-based approach, for the stochastic characterization and model identification of the stochastic errors related to low-cost MEMS-based inertial sensor errors. Raw inertial sensor measurements were collected using MTi-G-710 MEMS Inertial Measurement Unit (IMU) where the GMWM was utilized for estimating the coefficients of different random processes associated with inertial sensor residuals. Results demonstrated that highly complicated error model structures are identified for the tested IMU, which contradicts the common assumption of modeling such errors as a single 1
^st
order Gauss-Markov (GM) random process. In addition, the results showed that the GMWM is an efficient framework that could be used for estimating the parameters associated with complex structured random processes with the advantage of correlated residual characterization and identification.