Resource Allocation in Multi-UAV Communication Networks Using MADDPG Framework with Double Reward and Task Decomposition

Efficient resource allocation is essential for enhancing the performance of multi-UAV communication networks. However, the dynamic and complex nature of varying environments presents considerable challenges for solving resource allocation problems. In this letter, we propose a resource allocation algorithm built upon the Multi-Agent Deep Deterministic Policy Gradient (MADDPG) framework. Considering the difficulties in balancing overall and partial performance within the communication system, as well as the challenges in achieving cooperative optimization across different communication links, we propose MADDPG with Double Reward and Task Decomposition (MADDPG-DRTD) by using global reward and local task decomposition reward. Simulation experiments confirm the superior performance of MADDPG-DRTD algorithm. Compared to the exhaustive search strategy, our method exhibits lower complexity while maintaining slightly inferior performance.

I. Introduction

The incorporation of unmanned aerial vehicles (UAVs) into communication networks has emerged as a key area of research and development, driven by their potential to improve network coverage, capacity, and adaptability across various applications [1]. Recent advancements in 5 G and beyond-5G technologies have further highlighted the role of UAVs as key enablers of next-generation wireless communication systems. UAVs can provide on-demand connectivity, especially in scenarios where traditional infrastructure is unavailable or impractical, such as in disaster recovery, remote sensing, and large-scale events [2].