Resource Allocation in 5G Networks using Reinforcement Learning

Abstract

Device-to-Device (D2D) technology is becoming increasingly significant for boosting spectral utilization in emerging wireless networks. With the exponential rise in interconnected devices, Fifth-Generation (5G) networks must provide greater data rates accompanied by extremely low latency. To achieve this objective, this study focuses on optimizing total throughput for cellular and D2D links simultaneously within a cell using resource allocation methods, where several D2D pairs concurrently access a single cellular channel. Thus, we introduce an efficient Q-learning and Deep Q-Network (DQN) based channel allocation strategy for D2D communications operating alongside cellular networks. For developing the proposed Q-learning and DQN mechanism, an emulator was built to mimic the wireless network environment. The simulation results illustrate marked improvements in system throughput.