This thesis is concerned with determining the orientation of small Unmanned Aerial Vehicles(UAVs). To make commercial use of these aircraft in aerial surveying markets their attitude needs to be determined accurately and precisely throughout a survey flight.

Traditionally inertial sensors have been used on larger aircraft to estimate both position and orientation in combination with Global Navigation Satellite Systems (GNSS). High quality inertial sensors have many downsides when used on the small UAV. They are expensive, power hungry and often heavy. Inertial sensors are vulnerable to vibration, high acceleration, high rotation rate and jerk. All of these are present on the small UAV. This thesis identifies GNSS attitude determination as a potentially suitable alternative to inertial techniques.

Carrier phase GNSS attitude determination uses three or more GNSS receivers with antennas separated by a short baseline to estimate the orientation of the UAV. This technique offers low cost, high accuracy and drift-free attitude estimates. To be successfully used it requires removal of the biases present in the received GNSS signals and estimation of the integer cycle ambiguity present in the carrier phase measurement.

This thesis presents and examines the state of the art techniques for removing these biases and estimating an integer cycle ambiguity using a priori measurement of the interantenna distance. In this work a novel method is developed which uses this a priori baseline measurement to validate estimates of the carrier phase ambiguities.

In order to test these methods data has been gathered using low cost, commercially available GNSS receivers and antennas. This is the first work in which modern, low cost, GNSS equipment has been tested for use in attitude determination. It is found that the state of the art carrier phase GNSS attitude determination methods can provide an accurate attitude estimate for every set of measurements from the GNSS receivers.

However, a real UAV flight indicates that the low cost GNSS equipment does not track the GNSS signals throughout the flight. Signal outages, cycle slips and half cycle ambiguous carrier phase measurements occur due to rapid UAV manoeuvres. Having identified this problem this work goes on to replicate and quantify it through the use of a GNSS hardware simulator. Algorithms are then devised to increase the availability of the GNSS attitude solution throughout the tracking difficulties.

Complete GNSS signal tracking failures are overcome through the innovative use of kinematic and dynamic attitude models. Both types of model give an attitude solution throughout GNSS signal tracking problems without adding significant cost or weight to the system. When tracking of the GNSS carrier phase signal is possible, novel use of the carrier phase triple difference observable allows the attitude rate to be estimated even when the carrier phase measurements are half cycle ambiguous. It is shown that integer and half integer cycle slips can be removed from the measurement through the combination of the modelling and triple difference techniques.

The attitude output of both modelling and triple difference methods is used to resolve half cycle ambiguities and make full use of half cycle ambiguous data where previously it could not have been used. Success rates of up to 99.6% have been achieved for half cycle ambiguity resolution. As a result precise and accurate GNSS attitude solutions are available at nearly every epoch for which a carrier phase measurement is output by the GNSS receivers. When no measurement is available the attitude solution gracefully degrades over time.

This work makes reliable, accurate, low cost attitude determination possible on mini-UAVs.