An analytical framework for modeling distributed JRRM decision in cognitive networks

This paper deals with mobile-centered decision making in heterogeneous networks, where intelligent mobile terminals take autonomous decisions about the JRRM actions, consisting to connect to one of the available systems. This distributed decision-making is possible due to Q-learning algorithms implemented within the mobile terminals that enable them to profit from their past experience in order to enhance their subsequent decisions. We develop an original Markovian model that allows analyzing analytically the evolution of the Q-learning process and show how the the performance is enhanced until convergence.