In this thesis, we study wireless networks in which mobile terminals are free to choose their communication configuration. Theses configuration choices include access wireless technology, access point association, coding-modulation scheme, occupied bandwidth, power allocation, etc. Typically, these configuration choices are made to maximize some performance metrics associated to every terminals. Under the assumption that mobile terminals take their decisions in a rational manner, game theory can be applied to model the interactions between the terminals. Precisely, the main objective of this thesis is to study energy-efficient power control policies from which no terminal has an interest to deviate. The framework of stochastic games is particularly suited to this problem and allows to characterize the achievable utility region for equilibrium power control strategies. When the number of terminals in the network is large, we invoke mean field game theory to simplify the study of the system. Indeed, in a mean field game, the interactions between a player and all the other players are not considered individually. Instead, one only studies the interactions between each player and a mean field, which is the distribution of the states of all the other players. Optimal power control strategies from the mean field formulation are studied. Another part of this thesis has been focused on learning equilibria in distributed games. In particular, we show how best response dynamics and learning algorithms can converge to an equilibrium in a base station location game. For another scenario, namely a power control problem, we study the convergence of the best response dynamics. In this case, we propose a power control behavioral rule that converges to an equilibrium with very little information about the network.
