engleski

Algoritmi održavanja klastera u ad-hoc mrežama

The thesis gives an overview of key properties and design constraints for MANETs, and then focuses on clustering algorithms. Selected clustering algorithms found in literature are studied and compared by listing their advantages and disadvantages. Four new cluster maintenance algorithms have been proposed proposed in the thesis. They include: 1) the Cluster head centric maintenance algorithm, 2) the Destination-Sequenced Distance-Vector (DSDV) based maintenance, 3) the Connectivity-based maintenance algorithm, and 4) the Two-hop based maintenance algorithm. The algorithms have been implemented and evaluated in the OMNeT++ simulation environment. Three evaluation experiments have been designed in the simulation environment and run following an identical setup. In experiments I and II the percentage of nodes that change a cluster or form a new cluster is measured, and in experiment III the overall number of sent and received messages in the network is measured for each cluster maintenance algorithm. The proposed algorithms have also been compared against the periodic reclustering algorithm found in literature. The results show that the Cluster head centric maintenance algorithm gives the best overall performance. Finally, potential applications of clustered MANETs have been illustrated by using an e-health patient monitoring application as an example. Benefits of clusters are illustrated within use cases that require patient positioning or tracking, and data aggregation.

MANET; mobile ad-hoc wireless network; clustering; cluster maintenance; scalability; stability

Rad je napisan i obranjen na engleskom jeziku, sukladno proceduri propisanoj na Fakultetu elektrotehnike i računarstva Sveučilišta u Zagrebu.