| dc.description.abstract | Wireless mesh networks (WMNs) are networks whose main objective is to provide ubiquitous
and wireless connectivity to their clients through a set of mesh routers (MR). The WMNs
architecture can be classified into 3 groups depending on the functionalities of the nodes: infrastructure,
client WMNs, and hybrid wireless mesh networks as a combination of the above.
The client WMNs are networks whose management should be as spontaneous as possible. Nowadays,
spontaneous networks are emerging as a possible new paradigm of communication,
characterized by a strong self-organization and self-maintenance nature. The most common
cases are networks formed mainly by mobile devices carried by people. In such spontaneous
networks, the devices fulfill both the role of the end user interface, as well as being the traffic
router of their peers. Spontaneous networks are usually conformed by portable devices. When
these devices are associated with people moving from one place to another, they acquire the
human mobility attribute. Most of these networks with such mobility patterns present an organized
community structure.
Due to technological advance, mobile phones increasingly generate an area of interest for researchers,
in order to take advantage of real-time information, to benefit personal and environment
care. The combination of such smart devices with social relationships is a reason to incorporate
social characteristics into network design strategies. Due to the large number of devices that
can exist in these spontaneous networks, there are problems such as: overhead, high probability
of packet collision, interference, energy inefficiency, etc. Therefore, arises the need to implement
topology control mechanisms that take into account the nodes’s social interaction.
Under this context, this project proposes and evaluates three topology control schemes based
on centrality metrics in combination with community detection algorithms, for communitystructured
WMNs. Each of the evaluated schemes uses a different method for routers selection
which will form the network backbone. When evaluating the methods, it is concluded that the
Community-Aware Highest Betweenness Centrality Neighbor (C-A HCBN) method achieves the
best network connectivity with a reduced number of selected routers for all the mobility models under study. An analysis of the evolution of the reduced topologies is carried out. In all cases,
there is an average of less than 50% of nodes that are chosen to fulfill the functions of routers,
and the number of times that a node changes its state on average goes from 22 to 41 seconds
depending on the mobility model. C-A HCBN is also compared to a previous topology control
mechanism which doesn’t consider community structure. The simulation results with real-time
traffic (UDP) conclude that our method is better in terms of network performance and energy
efficiency.
Finally, we propose an additional topology control mechanism based on the minimum spanning
tree algorithm. This method manages to obtain smaller topologies than C-A HCBN However,
network performance with traffic load is affected by the reduced number of MR in this method.
The evaluation of the proposal is carried out by means of extensive simulations using the ns-3
software, a free licensed discrete event network simulator based on C ++, highly diffused and
used by the scientific community. The mobility traces are obtained by BonnMotion and SUMO,
which are tools for generation and mobility scenario analysis. Generation and visualization of
network graphs are obtained with Gephi. Other software that are used for data processing are
Octave and Python. | |
