الفهرس 
Wireless flow management model at a mobile host for MANETsOnly 14 pages are availabe for public view
 |  | from | 32 |  |  |
| | | from | 32 | | |
Abstract
Recently, Mobile Ad-Hoc Networks (MANETs) have received
much attention as an interesting category of wireless networks
that have many potential applications in which Quality-of-
Services (QoS) support is an essential aspect to be achieved
such as battlefield communication, disaster recovery, …etc.
However, managing link state information such as delay,
bandwidth, and cost is difficult because of the ever-changing
topology characteristics of the ad-hoc networks, limited
resources, and host’s mobility.
QoS routing is a key issue in designing these networks. A lot
of work has been done to support QoS in many types of wiredline
and cellular networks. Unfortunately, none of these works
can be applied directly in ad-hoc networks because of their
natures. Because the QoS routing problem is a hard
constrained distributed application with dynamically changing
constraints, and it is classified as NP-complete problem and it
lends itself to heuristic algorithms.
Since routing operations are very heavy tasks for routers using
traditional routing operations, the need of applying
evolutionary techniques such as genetic algorithms have
become increasingly interesting for research. Many routing
algorithms based on genetic algorithms are introduced.
In this thesis, an introduction to cellular and ad-hoc networks
is presented, and their different structures are discussed. QoS
routing as a main QoS component in ad-hoc networks is
introduced, its weighted graph model is presented, and unicast
and multicast routing problems are classified.
Because the multi-constrained routing process in ad-hoc
network environments is very exhaustive for routers (mobile
hosts). Finding routes under more than one constraint is an
NP-complete problem. The complexity of routing process in
ad-hoc network environment becomes extreme especially when the scalability of the network is taken into consideration.
Many heuristic techniques are studied to be used in the routing
process.
Ticket Based Probing routing algorithm is classified as one of
the best-known routing algorithms. This thesis proposes a
modified TBP algorithm that inherits the original ticket based
probing algorithm while reducing the message routing
overhead. The proposed algorithm uses hot and cold
termination strategies to achieve this reduction. Through
simulation, the proposed modified algorithm is found to have
the same success ratio of finding low cost routes but with
better message overhead ratio.
The second main contribution of this thesis is that, two genetic
algorithms for the ad-hoc QoS routing problems are proposed.
These two algorithms are different in data representation. The
first one is a simple genetic-based QoS unicast routing
algorithm in which a node-based encoding is applied for the
route representation. The second algorithm is a QoS-aware
genetic routing algorithm, which introduces a new link-based
encoding approach for the chromosome representation. The
simulation results and performance analysis show that the
application of genetic algorithms for the solution of QoS
routing problem is promising. The proposed algorithms are
found to have better QoS performance measures, namely;
success ratio and cost. The second algorithm achieves
acceptable success ration for various delay requirements.
The major achievements of this work could be highlighted as
follows:
1- A modified heuristic algorithm for QoS distributed
routing in mobile ad-hoc networks that reduces the
problem of routing message overhead is proposed.
2- A solution for the unicast source routing problem using
genetic algorithms is evaluated.
3- A QoS-Aware proposed genetic algorithm for unicast
source routing with a new proposed chromosomal
encoding is presented, with efficient performance.Information Systems Department