Fehmi BEN ABDESSLEM, UPMC-LIP6

Location :

Universite Pierre et Marie Curie - LIP6

104, avenue du President Kennedy
room 549
75016 Paris 

RER: Line C, station "Avenue du President Kennedy - Maison de Radio-France"
Metro (underground): Line 6, station: "Passy" 

http://www.lip6.fr/photos/infos/PlanKennedy750.jpg
http://maps.google.com/maps?q=104+Kennedy+Paris&t=h&ll=48.853472,2.281366&iwloc=addr

Abstract :

Examiners:

Isabelle GUERIN LASSOUS, Universite Claude Bernard (Lyon 1) 
Martin MAY, Thomson Paris Lab - ETH Zurich 
Nathalie MITTON, INRIA 
Franck ROUSSEAU, Grenoble INP
Artur ZIVIANI, LNCC 
Marcelo DIAS DE AMORIM, CNRS 
Serge FDIDA, Universite Pierre et Marie Curie (Paris 6)


Abstract:

Recent years have witnessed tremendous advances in wireless communications,
boosted by the expansion of the internet and mobile phone networks. More and
more users are connected everywhere and everytime. This trend for ubiquitous
wireless communications have emphasized the importance of self-organization and
distributed architectures for wireless networks. Although many algorithms,
protocols, and applications for self-organizing networks have been proposed by
the research community, most of them are not yet implemented and remain in the
simulation phase. However, the behavior and performance of wireless
communication algorithms are often different in real world conditions than in
simulation. The main reason for this phenomenon is that simulators cannot
accurately model signal propagation, user behavior, off-the-shelf hardware
performance, user mobility, and physical environment. Recently, some research
projects have acknowledged this issue by deploying experimental testbeds to
test and evaluate communication algorithms.


This thesis contributes in bridging the gap between theoretical solutions and
their real implementation on off-the-shelf hardware by enhancing the design
process of wireless communication algorithms and protocols. We argue that real
world conditions should be taken into account early in the design process of
wireless communication algorithms. To reach this goal, we advocate including a
prototyping phase in the design process of communication algorithms. We also
designed and implemented Prawn, a prototyping environment to help researchers
quickly and intuitively obtaining prototypes of their algorithm through simple
scripts, to be executed over real experimental testbeds. We show through a
number of case studies that Prawn provides an important first  insight into the
behavior of wireless algorithms in real conditions.


We also explored both sides of the design process of wireless systems by
designing a new wireless algorithm on the one side, and testing and measuring
the feasibility of wireless systems on a real testbed on the other side. First,
we designed a distributed cluster formation algorithm for large multi-hop
networks called Potential-Based Clustering (PBC). PBC does not require a global
knowledge of the network topology and provides a better control on cluster
sizes. Our cluster formation algorithm is based on the recursive propagation of
tokens according to the connectivity of nodes. We show through simulations that
with our approach, the average size of clusters is closer to the requested size
when compared to existing algorithms. Using Prawn, we easily implemented a
prototype of PBC running on a 400-node testbed. Second, as part of our
contribution to the design of a wireless peer-2-peer file exchange algorithm
for vehicular networks, we tested its feasibility in a real vehicular testbed
by performing extensive measurements. We investigate through real experiments
the characteristics of links formed by in-car nodes running off-the-shelf
wireless technologies such as IEEE~802.11(a/g) in ad hoc mode. We observe that
in-car nodes do show enough performance in terms of network capacity to be used
in a number of applications, such as file transfer in peer-to-peer
applications. Using Prawn, we also designed and prototyped a file transfer
mechanism over datagram packets for car-to-car communications, using forward
error control redundancy.


By improving the design process of wireless systems and bringing face to face
simulated algorithms and real experiments, this thesis contributes in bridging
the gap between the theoretical design of wireless systems and their
implementation on real testbeds.

Host :

Serge Fdida