Date of Award


Document Type


Degree Name

Doctor of Philosophy (PhD)


Department of Electrical and Computer Engineering

First Advisor

Meir Pachter, PhD


Sensor noise is an unavoidable fact of life when it comes to measurements on physical systems, as is the case in feedback control. Therefore, it must be properly addressed during dynamic system identification. In this work, a novel approach is developed toward the treatment of measurement noise in dynamical systems. This approach hinges on proper stochastic modeling, and it can be adapted easily to many different scenarios, where it yields consistently good parameter estimates. The Generalized Minimum Variance algorithm developed and used in this work is based on the theory behind the minimum variance identification process, and the estimate produced is a fixed point of a mapping based on the minimum variance solution. Additionally, the algorithm yields an accurate prediction of the estimation error. This algorithm is applied to many different noise models associated with three basic identification problems. First, continuous-time systems are identified using frequency domain measurements. Next, a discrete-time plant is identified using discrete-time measurements. Finally, the physical parameters of a continuous-time plant are identified using sampled measurements of the continuous-time input and output. Validation of the estimates is performed correctly, and the results are compared with other, more common, identification algorithms.

AFIT Designator


DTIC Accession Number