Betreuer: Markus Waelchli, Prof. Dr. Torsten Braun
Distributed event detection and event localization are
inherent tasks of many wireless sensor network applications. The inaccuracy of the sensed data and the
difficulty to determine event properties such as location, area diameter, etc. make
event detection and event localization challenging problems. In current approaches,
uncertainty is only barely considered and the detection and localization of events has
not yet been done in a fully distributed manner. In contrast, all data generated in
the neighborhood of an event is sent to a gateway and the localization of the event is
calculated there.
In this master thesis an efficient and fully distributed event
detection and event localization algorithm that avoids the drawback of increased data
traffic between the sensed area and a sink node shall be considered. Thereby, all sensors
in the neighborhood of an event derive a weight that determines the relevance
with which they sensed the event. In order to derive these weights, uncertainty will
have to be modeled. Therefore, a syntax for describing fuzzy rule sets and an appropriate
inference method will be proposed. Based on that information, a distributed election
algorithm is used that determines the relevant subset of sensors which are responsible
for handling the event further. The described approach as well as a simple centralized approach will
have to be implemented and evaluated in a simulation environment (Omnet++ or
JIST).
Start of Master's thesis: asap
Required skills: C/C++, basic knowledge in computer networks