The evolution of wireless communication and circuit technology has enabled the development of an infrastructure consists of sensing, computation and communication units that makes administrator capable to observe and react to a phenomena in a particular environment. The building block of such an infrastructure is comprised of hundreds or thousands of small, low cost, multifunctional devices which have the ability to sense compute and communicate using short range transceivers known as sensor nodes. The interconnection of these nodes forming a network called wireless sensor network (WSN). The low cost, ease of deployment, ad hoc and multifunctional nature has exposed WSNs an attractive choice for numerous applications. The application domain of WSNs varies from environmental monitoring, to health care applications, to military operation, to transportation, to security applications, to weather forecasting, to real time tracking, to fire detection and so on. By considering its application areas WSN can be argue as a traditional wired or wireless network. But in reality, these networks are comprised of battery operated tiny nodes with limitations in their computation capabilities, memory, bandwidth, and hardware resulting in resource constrained WSN. The resource constrained nature of WSN impels various challenges in its design and operations degrading its performance. On the other hand, varying numbers of applications having different constraints in their nature makes it further challenging for such resources constrained networks to attain application expectations. These challenges can be seen at different layer of WSNs starting from physical layer up to application layer. At routing layer, routing protocols are mainly concerned with WSN operation. The presence of these challenges affects the performance of routing protocols resulting in overall WSN performance degradation. The aim of this study is to identify the performance challenges of WSN and analyze their impact on the performance of routing protocols. For this purpose a thorough literature study is performed to identify the issues affecting the routing protocols performance. Then to validate the impact of identified challenges from literature, an empirical study has been conducted by simulating different routing protocols, taking into consideration these challenges and results are shown. On the basis of achieved results from empirical study and literature review recommendations are made for better selection of protocol regarding to application nature in the presence of considered challenges.

Vu les perspectives qu'ils offrent, les réseaux de capteur sans fil (RCSF) ont perçu un grand engouement de la part de la communauté de recherche ces dernières années. Les RCSF couvrent une large gamme d'applications variant du contrôle d'environnement, le pistage de cible aux applications de santé. Les RCSFs sont souvent déployés aléatoirement. Ce dispersement des capteurs nécessite que les protocoles de transmission utilisés soient résistants aux conditions environnementales (fortes chaleurs ou pluies par exemple) et aux limitations de ressources des nœuds capteurs. En effet, la perte de plusieurs nœuds capteurs peut engendrer la perte de communication entre les différentes entités. Ces limitations peuvent causer la perte des paquets transmis ce qui entrave l'activité du réseau. Par conséquent, il est important d'assurer la fiabilité des transmissions de données dans les RCSF d'autant plus pour les applications critiques comme la détection d'incendies. Dans cette thèse, nous proposons une solution complète de transmission de données dans les RCSF répondant aux exigences et contraintes de ce type de réseau. Dans un premier temps, nous étudions les contraintes et les challenges liés à la fiabilisation des transmissions dans les RCSFs et nous examinons les travaux proposés dans la littérature. Suite à cette étude nous proposons COMN2, une approche distribuée et scalable permettant de faire face à la défaillance des nœuds. Ensuite, nous proposons un mécanisme de contrôle d'erreur minimisant la perte de paquets et proposant un routage adaptatif en fonction de la qualité du lien. Cette solution est basée sur des acquittements implicites (overhearing) pour la détection des pertes des paquets. Nous proposons ensuite ARRP une variante de AJIA combinant les avantages des retransmissions, de la collaboration des nœuds et des FEC. Enfin, nous simulons ces différentes solutions et vérifions leurs performances par rapport à leurs concurrents de l'état de l'art. / Over the past decades, we have witnessed a proliferation of potential application domainsfor wireless sensor networks (WSN). A comprehensive number of new services such asenvironment monitoring, target tracking, military surveillance and healthcare applicationshave arisen. These networked sensors are usually deployed randomly and left unattendedto perform their mission properly and efficiently. Meanwhile, sensors have to operate ina constrained environment with functional and operational challenges mainly related toresource limitations (energy supply, scarce computational abilities...) and to the noisyreal world of deployment. This harsh environment can cause packet loss or node failurewhich hamper the network activity. Thus, continuous delivery of data requires reliabledata transmission and adaptability to the dynamic environment. Ensuring network reliabilityis consequently a key concern in WSNs and it is even more important in emergencyapplication such disaster management application where reliable data delivery is the keysuccess factor. The main objective of this thesis is to design a reliable end to end solution for data transmission fulfilling the requirements of the constrained WSNs. We tackle two design issues namely recovery from node failure and packet losses and propose solutions to enhance the network reliability. We start by studying WSNs features with a focus on technical challenges and techniques of reliability in order to identify the open issues. Based on this study, we propose a scalable and distributed approach for network recovery from nodefailures in WSNs called CoMN2. Then, we present a lightweight mechanism for packetloss recovery and route quality awareness in WSNs called AJIA. This protocol exploitsthe overhearing feature characterizing the wireless channels as an implicit acknowledgment(ACK) mechanism. In addition, the protocol allows for an adaptive selection of therouting path by achieving required retransmissions on the most reliable link. We provethat AJIA outperforms its competitor AODV in term of delivery ratio in different channelconditions. Thereafter, we present ARRP, a variant of AJIA, combining the strengthsof retransmissions, node collaboration and Forward Error Correction (FEC) in order toprovide a reliable packet loss recovery scheme. We verify the efficiency of ARRP throughextensive simulations which proved its high reliability in comparison to its competitor.