Date of Graduation


Document Type


Degree Type



Statler College of Engineering and Mineral Resources


Mechanical and Aerospace Engineering

Committee Chair

Marcello R. Napolitano.


Navigation-grade inertial sensors are often too expensive and too heavy for use in most Small Unmanned Aerial Vehicle (SUAV) systems. Low-cost Micro-Electrical-Mechanical-Systems (MEMS) inertial sensors provide an attractive alternative, but currently do not provide an adequate navigation solution alone due to the presence of sensor bias. Toward addressing this problem, this research focuses on the development and experimental evaluation of sensor fusion algorithms to combine partially redundant information from low-cost sensor to achieve accurate SUAV attitude estimation. To conduct this research, several sets of SUAVs flight data that include measurements from a low-cost MEMS based Inertial Measurement Unit, a Global Positioning System receiver, and a set of low-grade tri-axial magnetometers are used to evaluate a variety of algorithms. In order to provide a baseline for performance evaluation, attitude measurements obtained directly with a high-quality mechanical vertical gyroscope are used as an independent attitude 'truth'. In addition, as a part of this project, a custom SUAV avionics system was developed to provide a platform for fault-tolerant flight control research.;The overall goal of this research is to provide high-accuracy attitude estimation during nominal sensor performance conditions and in the event of sensors failures, while using only low-cost components. To achieve this goal, this study is carried out in three phases. The specific aim of the first phase is to obtain high-accuracy under nominal sensor conditions. During this phase, two different nonlinear Kalman filtering methods are applied to various sensor fusion formulations and evaluated with respect to estimation accuracy over diverse sets of flight data. Next, during the second phase, sensor fusion based calibration techniques are explored to further enhance estimation accuracy. Finally, the third phase of the study considers the design of a sensor fusion attitude estimation architecture that rejects sensor failures and supports a graceful loss of performance in the event of sensor failures. This algorithm is based on a nonlinear information filter and features a novel Failure Detection, Identification, and Accommodation (FDIA) approach. During this phase, it is shown that the sensor fusion algorithm handles both abruptly occurring large magnitude sensor failures, as well as gradually growing small magnitude sensor failures.