Centralized and decentralized multi-user schemes for wireless communications

As wireless systems grow, new challenges are emerging. The exact location of the intelligence is of paramount importance: a central controller (network centric) or the users (user centric)? The performance analysis of multiuser systems is done by novel use of sophisticated methodologies: random matrix theory and game theory. The purpose is to deliver optimal quality of service to users, under constraints like energy consumption and limited knowledge of the environment. First, random matrix theory is used to optimize large cellular networks, for which simulations involve a huge number of parameters. The self-averaging effect of random matrices enables to elegantly single out parameters of interest in asymptotic systems, when both the number of chips and of users grow large with fixed ratio. Although asymptotic, results give accurate predictions of the finite size behavior, as shown by simulations. The performance analysis of centralized systems gives a bound on performance. Second, game theory is applied to distributed multiuser schemes. Game theory studies interactions among selfish players who reason strategically in order to take rational decisions. This is of considerable interest for the deployment of self-organizing networks. Some subfields are particularly promising. Correlated games enable a simple coordination mechanism between players, while evolutionary games provide additional properties of robustness of equilibrium strategies. In the asymptotic regime, non-atomic games study interactions of dense populations where the behavior of a single individual has a negligible impact on the welfare of the population as a whole.