7. Exploiting User Demand Diversity: QoE Game and MARL Based Network Selection

This chapter studies distributed network selection for multiple user cases with heterogeneous demand. The key challenge is low system efficiency due to user competition. Motivated by the fact that the ultimate goal of communications is to serve users with personalized demand, we introduce a new concept of user demand diversity gain. This gain derives from the elaborate matching between user demand and radio resource, which cannot be directly attained in the existing throughput-centric optimization due to users’ blindness in maximizing throughput. Aiming at obtaining this gain, we propose the user demand centric optimization, where users seek to maximize QoE. To model this problem, we propose a novel game formulation, QoE game. The properties of QoE game and QoE equilibrium and the definition of user demand diversity gain are presented. Two distributed QoE equilibrium learning algorithms, stochastic learning automata (SLA) based algorithm and trail and error (TE) based algorithm, are designed to achieve QoE equilibrium. Simulation results validate the existence of user demand diversity gain and the effectiveness of the proposed learning algorithms in improving the system efficiency and QoE fairness.