Analysis and design of cognitive wireless multihop networks

Abstract

In the last decade, the deployment of wireless local area networks (WLANs) based on the IEEE 802.11 standard [1] has grown signi cantly, mainly due to the use of unlicensed frequency bands. It also became the de facto standard for the socalled wireless mesh networks (WMNs) [2{4]. In this context, a typical application of this type of wireless solutions is to provide Internet access in suburban and rural areas. This kind of networks was useful for the deployment of Plan Ceibal [5], the nationwide implementation in Uruguay of the novel one-to-one educational model, which was the main motivation to carry on this thesis. The issue is how to optimize these networks to meet the growing tra c demands and the new requirements imposed by the future applications. In this regard, we start this thesis presenting a characterization and statistical model of a WLAN, on the one hand analyzing the demand in the educational context of Plan Ceibal, and on the other hand estimating the capacity of 802.11-based wireless links from measurements of the physical layer. In the second part of this work, we propose a new routing and forwarding scheme for multihop wireless networks, based on the development of a statistical model of the links' queues, learned from live network measurements. We address the problem of deciding the most suitable path for the packet flows between each source-destination pair. We seek an optimal solution, capable of balancing the traffic load based on the resources available in each link. We present a suitable algorithm that solves the optimization problem posed in a distributed manner. Several simulations in di erent scenarios were performed to verify the performance of the proposed method, and also to compare with other schemes. In all the simulations, independently of the topology size, we observed a quick adaptation of the proposed algorithm to traffic changes and also an stable operation, avoiding the routing oscillations of the routing method included in the 802.11s standard, already noticed before by [6, 7]. Finally, the last part of the thesis is founded on the cognitive radio networks¡ (CRNs) paradigm [8]. In this case we propose a novel robust spectrum allocation mechanism that takes advantage of the free spaces in licensed bands to expand the network resources (e.g. spectrum holes in TV bands). The introduced robust method was evaluated by several simulations for different network topologies. The results show that in all cases the robust approach ensures compliance with the effective capacity required on each wireless link with high probability. The additional spectrum for this robustness is below 35% more than the optimality bound, given by the case of knowing in advance the primary users activity.