Dynamic Actor-Critic: Reinforcement Learning Based Radio Resource Scheduling for LTE-Advanced

Abstract

This paper proposes, an Actor-Critic Reinforcement learning based radio resource scheduling policy in downlink Transmission for Long Term Evaluation Advanced (LTE-A) radio resource technology. The scheduling technique uses the neural network (NN) based actor critic architecture in order to propose proper scheduling rules at each Transmission Time Interval (TTI). The objective is to improve system capacity, system throughput and spectral efficiency. NN based Actor-critic Reinforcement learning is proposed to accomplish the resource scheduling efficiently by maintaining best QoS capabilities and user fairness. The simulation results indicate that the proposed method achieves desired throughput and increased convergence capability.