Abstract:
Multi-radio Multi-channel (MRMC) Wireless Mesh Networks (WMNs) have
made rapid progress in recent years to become a preferred option for end
users due to its reliability, scalability and extending the network connectiv-
ity on the last mile. Although WMNs have already been deployed but still
the capacity of WMNs is limited due to non-coordinated interference (nCO)
among channels. To minimize non-coordinated interference among channels
and maximize network capacity; channel assignment has always been a key
issue in WMNs. Assigning IEEE 802.11b non-overlapping or Orthogonal
Channels (OCs) minimize channel interference but they constraint network
capacity as they are limited in number and lead to co-channel interference.
On the other hand channels whose spectrum interferes with each other are
considered as partially overlapping channels in IEEE 802.11b. These partially
overlapping channels are not used for transmission as they result in transmis-
sion losses due to adjacent channel interference. Recent studies have shown
that Partially Overlapping Channels (POCs) are not harmful and they can be
used to further improve network performance and can utilize IEEE 802.11b
spectrum more e ciently. In this thesis, we propose an optimization model
that maximizes network capacity and minimizes non-coordinated interference
using both non-overlapping and partially overlapping channel assignment in
MRMC-WMNs. We also propose an infrastructure heuristic channel assign-
ment algorithm POCA (Partially Overlapping Channel Assignment) where
a central node keeps and distributes all the channel assignment information.
The proposed model assigns both non-overlapping and partially overlapping
channels and nds the optimum channel assignment strategy. Simulation
results show that POC assignment performs 17% better than traditional
non-overlapping channel assignment in sparse WMN topologies where the
non-coordinated interference is high. For dense WMN network environments,
where the non-coordinated interference is not high, POC performs 9% better
than OC assignment.
