Detection of wireless signals from an unmanned aerial vehicle for search and rescue.
Thesis DisciplineComputer Science
Degree GrantorUniversity of Canterbury
Degree NameMaster of Science
This thesis investigates the application of several spectrum sensing algorithms to the problem of detecting disaster victims after a large scale disaster using a unmanned aerial vehicle (UAV) based detector. The spectrum sensing algorithms are: the energy detector, the matched filter detector and two cyclostationary feature detector algorithms ,  Each of the detectors were implemented as constant false alarm rate (CFAR) tests in order to guarantee an upper bound for false positive detections and all simulations were performed with a false probability rate of 5% and 1%.
The ultimate goal of the greater research effort is to develop a means of detecting the wireless signals transmitted by the body area network (BAN) of a victim before attempting to connect and retrieve health data from the fitness tracker devices connected to the BAN in order to help triage the victim. In consideration of this greater goal, performance curves for the four spectrum sensing algorithms were plotted and compared against each other to determine which approach would be best for the end goal.
After considering the computational complexity of the algorithms and their resilience to low signal to noise ratio (SNR) conditions, it was found that the matched filter detector is the most resilient against high noise, but as this method requires a static signal kernel to match against, it is unable to detect some signal types. As a result, the cyclostationary feature detector described by Lundén is also valuable due to its low computational complexity and ability to detect signals such as those transmitted by mobile devices using Long-Term Evolution (LTE).
In addition to the main goal of the project, a similar but unrelated project regarding the development of a UAV based detector for locating very high frequency (VHF) wildlife tracking beacons to assist with conservation work. The developed detector has been tested in the field for two several month long trips to assist with active research, and been shown to greatly reduce the amount of man-hours required to locate the wildlife tracking beacons as compared to previous approaches.