A Nash-Stackelberg Fuzzy Q-Learning Decision Approach in Heterogeneous Cognitive Networks

Motivated by the fact that when selfish users choose their policies independently without any coordination mechanism, Nash equilibria could result in a network collapse, we develop in this paper a hierarchical distributed learning framework for decision-making in heterogeneous cognitive networks. We introduce the Nash-Stackelberg fuzzy Q-learning, with the network as leader that aims at maximizing its utility (revenue) and the mobiles as followers that have their individual objectives (maximizing their QoS). We validate our results through extensive simulations of the algorithm in a practical setting of a geographical area covered by a global HSDPA and 3G LTE system that serves both streaming and elastic traffic.