Joint Interference Alignment and Power Control for Dense Networks via Deep Reinforcement Learning

Abstract

This letter proposes a joint interference suppression scheme in heterogeneous networks (HetNets) with dense small cells (SCs) and users. Different from the majority of existing studies, we adopt the co-tier intra-cell interference alignment (IA), while the co-tier inter-cell and cross-tier interference is suppressed by centralized power control in the macro base station (MBS). Specifically, the power control problem is modeled as a Markov Decision Process (MDP) with the aim of maximizing the sum spectrum efficiency. Considering the exponential growth of the output layer neurons faced by general deep reinforcement learning (DRL) algorithms, we propose a deep deterministic policy gradient (DDPG)-based algorithm to solve the problem. Simulation results demonstrate that the proposed algorithm is able to achieve better performance and wider application scope comparing with existing algorithms.