Date of Award


Access Type

Thesis - Open Access

Degree Name

Master of Science in Mechanical Engineering


Mechanical Engineering

Committee Chair

Dr. Hever Moncayo

First Committee Member

Dr. Marc Compere

Second Committee Member

Dr. Richard Prazenica


Since the 1990’s, there has been increased focus on creating navigation systems for small unmanned systems, particularly small unmanned aerial systems (SUAS). Due to size, weight, and cost restrictions, compared to larger more costly manned systems, navigation systems for SUAS have evolved to be quite different from the proven systems of the past. Today, there are many solutions for the problem of navigation for SUAS. The problem has now become choosing the most fitting navigation solution for a given application/mission. This is particularly true for evaluating solutions that are fundamentally different.

This research analyses the performance and sensitivity of four sensor fusion solutions for attitude estimation under multiple simulated flight conditions. There are three different hardware configurations between the four estimators. For this reason, each estimator is tuned to be experimentally optimal, as to provide a fair comparison between different estimators. With each estimator tuned to its highest performance, the estimators are compared based on their sensitivity to tuning error, sensor bias, and estimator initialization error. Finally the estimators' accuracy performances are directly compared.

This thesis also provides methods to tune different configuration estimators to their individual best performances. These methods show that choosing tuning parameters based on sensor noise covariance, as is typically done in research, does not produce optimal performance for all estimator formulations. After comparing multiple sensitivity and performance properties of the estimators, observations are provided regarding the efficacy of the analyses, including the applicability of the metrics used to determine performance. Some metrics where shown to be misleading for particular estimators or analyses. Ultimately, guidance is given for choosing performance metrics capable of comparing different solutions.