Energy-Efficient Communication Networks via Multiple Aerial Reconfigurable Intelligent Surfaces: DRL and Optimization Approach

TL;DR

This paper proposes an energy-efficient communication system using multiple UAV-mounted reconfigurable intelligent surfaces, employing DRL and optimization techniques to enhance coverage and efficiency in next-generation wireless networks.

Contribution

It introduces a joint optimization framework for deploying and controlling multiple ARISs in UAV-assisted networks, combining DRL and heuristic algorithms to solve a complex NP-hard problem.

Findings

Proposed algorithms significantly improve energy efficiency and coverage.

Simulation results validate the effectiveness of the joint optimization approach.

The system outperforms baseline methods in various scenarios.

Abstract

In the realm of wireless communications in 5G, 6G and beyond, deploying unmanned aerial vehicle (UAV) has been an innovative approach to extend the coverage area due to its easy deployment. Moreover, reconfigurable intelligent surface (RIS) has also emerged as a new paradigm with the goals of enhancing the average sum-rate as well as energy efficiency. By combining these attractive features, an energy-efficient RIS-mounted multiple UAVs (aerial RISs: ARISs) assisted downlink communication system is studied. Due to the obstruction, user equipments (UEs) can have a poor line of sight to communicate with the base station (BS). To solve this, multiple ARISs are implemented to assist the communication between the BS and UEs. Then, the joint optimization problem of deployment of ARIS, ARIS reflective elements on/off states, phase shift, and power control of the multiple ARISs-assisted communication system is formulated. The problem is challenging to solve since it is mixed-integer, non-convex, and NP-hard. To overcome this, it is decomposed into three sub-problems. Afterwards, successive convex approximation (SCA), actor-critic proximal policy optimization (AC-PPO), and whale optimization algorithm (WOA) are employed to solve these sub-problems alternatively. Finally, extensive simulation results have been generated to illustrate the efficacy of our proposed algorithms.