Optimal Identification of IMU Drift Models

This talk deals with the identification via Maximum Likelihood Estimation of the variances of process and measurement noises for a Wiener process drift model and, additionally, the time constant of an Ornstein-Uhlenbeck (Gauss-Markov of order 1) model together with its noise variances. These are the most common models for Inertial Navigation Units (gyroscopes and accelerometers) drift models. The CRLB for the estimated parameters is also evaluated and it is shown the MLE yields statistically efficient estimates, i.e., it extracts all the information from the noisy data - it is optimal - there can be no better estimator. The MLE for these two models are illustrated on simulated as well as real data. A statistical procedure for model validation/selection is also presented.