Learning-Based Resource Partitioning in Heterogeneous Networks With Multiple Network Operators

In heterogeneous network, it is important to mitigate cross-tier interference. Resource partitioning is the one of solution to reduce interference. However, most of studies on partitioning assumed that there was only one network operator to cooperate with. In this letter, we study the network selection problem of small access points in heterogeneous networks, which are provided by multiple network operators. We model the multiple network operators scenario as a congestion game. To solve the equilibrium point of suggested game, we analyze some features of proposed model such as potential game property, smoothed best response dynamics and logit equilibrium. Then, we propose a reinforcement learning algorithm that can reach logit equilibrium in distributed way. Moreover, we also suggest the adjustment of learning parameters to enhance adaptability. By means of simulations, it is shown that proposed algorithm has near-optimal performance in view of throughput, fairness and adaptability.