Resource Allocation in OFDMA Systems Using Reinforcement Learning

Long Term Evolution (LTE) makes use of Orthogonal Frequency Division Multiplexing (OFDM) to achieve high downlink data rates. Based on channel conditions of users, a resource scheduler can use dynamic resource allocation to provide best resource to users and optimize system performance. Proportional Fair (PF) scheduling is a popular resource allocation algorithm that tries to maximize total throughput while providing all users at least a minimal level of service. Recent advances in reinforcement learning (RL) has encouraged efforts to formulate resource allocation as a RL problem. In this study, Deep Deterministic Policy Gradient technique of RL is used to design RL-based resource scheduler and its performance is compared with that of the PF scheduler.