Multi-Agent Deep Reinforcement Learning for Distributed Resource Management in Wirelessly Powered Communication Networks

TL;DR

This paper introduces a multi-agent deep reinforcement learning approach for distributed resource management in wireless powered communication networks, enabling decentralized decision-making with performance comparable to centralized methods.

Contribution

It proposes a novel MADRL framework where each access point independently optimizes resource allocation based on local information, reducing reliance on global data.

Findings

Achieves performance comparable to centralized algorithms.

Enables decentralized control in multi-cell WPCNs.

Demonstrates effectiveness through numerical simulations.

Abstract

This paper studies multi-agent deep reinforcement learning (MADRL) based resource allocation methods for multi-cell wireless powered communication networks (WPCNs) where multiple hybrid access points (H-APs) wirelessly charge energy-limited users to collect data from them. We design a distributed reinforcement learning strategy where H-APs individually determine time and power allocation variables. Unlike traditional centralized optimization algorithms which require global information collected at a central unit, the proposed MADRL technique models an H-AP as an agent producing its action based only on its own locally observable states. Numerical results verify that the proposed approach can achieve comparable performance of the centralized algorithms.