Avec le succès des communications sans fil, il devient possible d'accéder au réseau partout et à tout moment sans avoir recours à connecter physiquement les appareils communicants à une infrastructure. Les nœuds (ordinateurs portables, smartphones, etc) peuvent analyser les différents canaux radio afin de pouvoir s'associer à un réseau sans fil disponible (station de base, point d'accès, etc.). Un avantage indéniable de ses technologies sans fil est la possibilité d'être mobile tout en restant connecté. Cependant, la mobilité est une tâche difficile à gérer car elle doit être abordée à différentes couches pour être transparente aux utilisateurs. Dans les MANET (Mobile Ad hoc Network), les protocoles de routage utilisent des métriques pour sélectionner les meilleures routes. Les métriques peuvent refléter la qualité de la liaison sans fil et aider à gérer la mobilité.Mais, un retard important entre l'estimation des métriques et leur inclusion dans le processus de routage rend cette approche inefficace.Les travaux de cette thèse s'intéressent à la proposition de nouvelles méthodes de calcul des métriques de routage pour gérer le problème de la mobilité dans les réseaux ad hoc. Les nouvelles métriques doivent refléter la qualité du lien et être sensibles à la mobilité en même temps. Nous considérons les métriques classiques, en particulier ETX (Expected Transmission Count) et ETT (Expected Transmission Time). Nous introduisons de nouvelles méthodes pour anticiper les valeurs de ces métriques en utilisant des algorithmes de prédiction. Nous utilisons une approche Cross layer, qui permet l'utilisation conjointe de l'information à partir des couches 1, 2 et 3. La validation de nouvelles méthodes de calcul des métriques de routage nécessite une évaluation au travers d'un véritable banc d'essai. Nous avons donc également mis en œuvre les nouvelles métriques de routage dans un testbed afin d'évaluer et de comparer leurs performances avec les métriques classiques. / With the success of wireless communications, it becomes possible to access the network anywhere at any time without the need for physically connect communicating devices in an infrastructure. The nodes (laptops, smartphones, etc.) can analyze different radio channels to be able to associate with an available wireless network (base station, access point, etc.). An undeniable advantage of wireless technologies is the ability to be mobile while staying connected. However, mobility is difficult to manage because it must be addressed at different layers to be transparent to users. In MANET (Mobile Ad hoc Network) routing protocols use metrics to select the best routes. The metric can reflect the quality of the wireless link and help manage mobility.But a significant delay between the estimate metrics and their inclusion in the routing process makes this approach ineffective.The work of this thesis are interested in proposing new methods for calculating routing metrics to manage the problem of mobility in ad hoc networks. The new metrics should reflect the quality of the link and be sensitive to the mobility simultaneously. We consider the classical metrics, particularly ETX (Expected Transmission Count) and ETT (Expected Transmission Time). We introduce new methods to predict the values of these metrics using prediction algorithms.We use a cross layer approach, which allows the joint use of information from layers 1, 2 and 3. The validation of new methods for calculating routing metrics requires evaluation through a real bench test. So we also implemented new routing metrics in a testbed to assess and compare their performance with classical metrics.

This thesis contains a theoretical basis for MANET networks. The main focus of the thesis is principles of these networks, their routing protocols and especially on Ad hoc On-Demand Distance Vector (AODV), implementation of this protocol and implementation of new mechanis for peer selection. Thesis contains ETX metric basic and implementation of this metric to AODV protocol. There is a demonstration of simulation of AODV protocol and simulation of AODV-ETX protocol. Simulations are run in Network Simulator 3. AODV and AODV-ETX comparasion are included in thesis.

In the past few years Wireless Mesh Networks (WMNs) have developed as a promising technology to provide flexible and low-cost broadband network services. The Expected Transmission Count (ETX) routing metric has been put forward recently as an advanced routing metric to provide high QoS for static WMNs. Most previous research in this area suggests that ETX outperforms other routing metrics in throughput and efficiency. However, it has been determined that ETX is not immune to load sensitivity and route oscillations in a single radio environment. Route oscillations refer to the situation where packet transmission switches between two or more routes due to congestion. This has the effect of degrading performance of the network, as the routing protocol may select a non optimal path. In this thesis we avoided the route oscillation problem by forcing data transmission on fixed routes. This can be implemented in the AODV (Ad hoc On-demand Distance Vector) protocol by disabling both error messages and periodic updating messages (the HELLO scheme). However, a critical factor for our approach is that ETX must determine a high quality initial route in AODV. This thesis investigates whether the ETX metric improves initial route selection in AODV compared to the HOPS metric in two representative client-server applications: the Traffic Control Network (TCN) and the Video Stream (VS) network. We evaluate the ETX and HOPS metrics in a range of scenarios which possess different link qualities and different traffic loads. We find the ETX metric greatly improves initial route selection in AODV compared to the HOPS in the network in which only single flow exists. For networks in which there are multiple simultaneous flows, ETX behaves similar to HOPS in initial route selection. Based on these results, we find the solution of route stabilization to route oscillations in the context of ETX is only useful in the single flow case. To address this problem, we propose a modified solution of repeatedly broadcasting RREQ (Route Request) packets. Simulation results show that our modified solution allows ETX to be useful in the initial route selection in both single flow and multiple simultaneous flows cases.

In the past few years Wireless Mesh Networks (WMNs) have developed as a promising technology to provide flexible and low-cost broadband network services. The Expected Transmission Count (ETX) routing metric has been put forward recently as an advanced routing metric to provide high QoS for static WMNs. Most previous research in this area suggests that ETX outperforms other routing metrics in throughput and efficiency. However, it has been determined that ETX is not immune to load sensitivity and route oscillations in a single radio environment. Route oscillations refer to the situation where packet transmission switches between two or more routes due to congestion. This has the effect of degrading performance of the network, as the routing protocol may select a non optimal path. In this thesis we avoided the route oscillation problem by forcing data transmission on fixed routes. This can be implemented in the AODV (Ad hoc On-demand Distance Vector) protocol by disabling both error messages and periodic updating messages (the HELLO scheme). However, a critical factor for our approach is that ETX must determine a high quality initial route in AODV. This thesis investigates whether the ETX metric improves initial route selection in AODV compared to the HOPS metric in two representative client-server applications: the Traffic Control Network (TCN) and the Video Stream (VS) network. We evaluate the ETX and HOPS metrics in a range of scenarios which possess different link qualities and different traffic loads. We find the ETX metric greatly improves initial route selection in AODV compared to the HOPS in the network in which only single flow exists. For networks in which there are multiple simultaneous flows, ETX behaves similar to HOPS in initial route selection. Based on these results, we find the solution of route stabilization to route oscillations in the context of ETX is only useful in the single flow case. To address this problem, we propose a modified solution of repeatedly broadcasting RREQ (Route Request) packets. Simulation results show that our modified solution allows ETX to be useful in the initial route selection in both single flow and multiple simultaneous flows cases.

The use of Internet of Things systems has increased to meet the need for smart systems in various fields, such as smart homes, intelligent industries, medical systems, agriculture, and the military. IoT networks are expanding daily to include hundreds and thousands of IoT devices, which transmit information through other linked devices to reach the network sink or gateway. The information follows different routes to the network sink. Finding an ideal routing solution is a big challenge due to several factors, such as power, computation, storage, and memory limitation for IoT devices. In 2011, A new standardized routing protocol for low-power and lossy networks was released by the Internet Engineering task force (IETF). The IETF adopted a distance vector routing algorithm for the RPL protocol. RPL protocol utilizes the objective functions (OFs) to select the path depending on diffident metrics.These OFs with different metrics must be evaluated and tested to develop the best routing solution.This project aims to test the performance of standardized RPL objective functions in a simulation environment. Afterwards, a new objective function with a new metric will be implemented and tested in the same environmental conditions. The performance results of the standard objective functions and the newly implemented objective function will be analyzed and compared to evaluate whether the standard objective functions or the new objective function is better as a routing solution for the IoT devices network.