Soutenance de doctorat en science de Mr DAAS Mohamed Skander

Date : 27 Janvier 2020
Lieu : Salle de soutenance à la faculté des NTICs
Organisé par l'équipe SCAL
Mots clés
Mobile Ad hoc Networks MANETs VANETs Protocols Cross-layer Routing Swarm Intelligence Stigmergy Design of Experiments Local Maximum.

Given the emergence of different types of communication technologies that
make communication possible from smartphones to vehicles and given the specific characteristics
of mobile ad hoc networks, routing in this type of network is even more
constrained by several external factors. However, existing protocols need to be re-designed
and optimized to create new optimal architectures. As a result, the use of swarm intelligence
and design of experiments seems to be a real challenge.
When exploring the state of the art on swarm intelligence, we were able to improve
an approach of swarm intelligence called (BFO). In order to understand the evolution
of the different performances of routing protocols in MANET networks according
to the different influencing parameters, we used design of experiments to model and
analyze the different performance metrics of such protocols. The results showed that the
studied performances follow quadratic models. This type of models can be exploited for
the analysis and the improvement of routing protocols.
In order to improve the routing performances in MANET networks, we have proposed
two mechanisms inspired by swarm intelligence and we have used the design of
experiments as a modeling and optimization tool. Proposed mechanisms and techniques
are tested on a recent cross-layer geographic routing protocol for VANET networks called
CLWPR. First, we used Stigmergy, a form of implicit communication inspired by ant
behavior to reduce the size of routing packets. Secondly, to reduce the end-to-end delay,
we adopted a behavior inspired by the common movement of birds. Thirdly, for the same
purpose of reducing delay, we proposed to adapt the maximum packet caching time in
real time based on quadratic models using response surface methodology (RSM). The
results show that the employed mechanisms are able to reduce the size of the routing
packets and to reduce the delay without deteriorating the delivery ratio.