Allocation opportuniste de spectre pour les radios cognitives

Since a few years, the demand for wireless services and technologies has increased. According to several market studies this tendency is forecasted to follow a sustained growth in the years to come. However, on the one hand the new wireless services and technologies require increasing spectrum resources to work, while on the other hand several countries already suffer from a "spectrum scarcity" in their regulatory spectrum allocation charts. Thus, the question becomes: either "where to put in the spectrum band the new technologies?" or "how to rearrange the existing spectrum allocation model to fit new technologies?" Indeed, several studies have shown that the spectrum scarcity is much more in the allocation than in the use: not all the allocated spectrum is really used all the time and everywhere (existence of "spectrum holes"). Accordingly, to fully benefit from the potential of new radio access technologies, a new regulatory framework must be defined for spectrum allocation and use. At the same time, smarter and more flexible algorithms are required for managing spectrum use between users. Our objective in this thesis is to propose innovative technical algorithms of dynamic spectrum allocation. Assuming smarter devices (more intelligence and environment-awareness are included within user terminals), as in the cognitive radio context, we study bottom-up distributed algorithms of opportunistic spectrum access. We are interested in studying the spectrum allocation problem in scenarios of several distributed users sharing (spectrum pool) several channels. We propose a new distributed algorithm at the MAC layer for UpLink (UL) spectrum access. It can be applied in OFDM-based systems such as, for example, WLAN systems. In our bottom-up approach users collectively and opportunistically negotiate their spectrum allocation, using a collaborative contention mechanism. This algorithm adapts even under a variable available spectrum resource (bandwidth and position). This is particularly suitable in dynamic environments and when the number of users is important. Our OFDMA algorithm uses, in a wireless context, an adaptation of the swarm intelligence meta-heuristic, a method inspired from the study of social insects (e.g.: ants, bees, wasps, termites). It allows flexibility, robustness and dynamicity in the spectrum allocation. Our results show the real potential of such distributed collaborative methods for tackling the spectrum allocation problem. The complexity of a real wireless network is recreated at the global level, by means of simple (controlled) local interactions. Such a bottom-up approach requires completely new modeling rules/habits, which is a real challenge to overcome, but whose final benefit can be huge for next generations of wireless systems.