Power Control Based on DRL Algorithm for D2D-Enabled Networks

The problem of power control in the uplink network of cellular users communicating with Device-to-Device (D2D) is mainly studied. Since cellular users and D2D users share spectrum resources, several serious interference will be caused by them. Reasonable measures are taken to control the interference caused by the sharing spectrum resources, otherwise that influences the quality of service (QoS) of cellular users. The reduction of entire system interference can be achieved by power control, so a method of power control based deep reinforcement learning (DRL), namely Asynchronous Advantage Actor Critic (A3C) Algorithm, is proposed. At the same time, the spectrum resources utilization of the system is improved and QoS of cellular users is guaranteed. The simulation results prove rationality of the proposed algorithm, and have better convergence performance than the traditional DRL algorithm.