Multi-Uav Trajectory Planning for Ris-Assisted Swipt System Under Connectivity Preservation

Abstract

Information transmission and energy supply are two requisites for low-power Internet-of-things (IoT) devices to work effectively, especially in geographically constrained or disaster-affected areas. As an emerging technology, simultaneous wireless information and power transfer (SWIPT) ensures uninterrupted communication through supplying power to exclude battery-depleted situation. Owing to the  high maneuverability, unmanned aerial vehicle (UAV) has enabled SWIPT a new paradigm by providing flexible and on-demand service. The connectivity preservation among UAVs when they move to conduct tasks is an inescapable issue, for their powerful online decision-making requires information sharing. However, this issue is omitted in existing UAV-aided wireless communication literatures. In this paper, we propose a novel connectivity preserved multi-agent deep reinforcement learning algorithm (CP-MARL) to solve the trajectory planning problem in complex environments, where multi UAVs cooperatively provide SWIPT service for ground users (GUs) while maintaining information exchanging. Moreover, the fairness of service received among GUs is also considered to avoid the suboptimal system performance caused by partial resource allocation. Besides, reconfigurable intelligent surface (RIS) is adopted to redirect the signal to achieve line-of-sight (LoS) channel. Under this deployment, we intend to maximize the fair energy efficiency by jointly optimizing the  trajectory of UAVs, the phase shift of RIS, power splitting (PS) ratio and the association relationship between UAVs and GUs. The formulated problem is non-convex in its original form and the proposed CP-MARL achieves online near-optimal solutions. Numerical results verify that the designed CP-MARL algorithm effectively improves the energy-efficiency performance of the system, reaching 40.22\% performance gain versus its counterpart where connectivity is not guaranteed, and the training process is more stable. Additionally, the superiority of incorporating RIS in this system is also validated by the 27.31\% energy-efficiency improvement.

Keywords: UAV, trajectory, MARL, RIS, SWIPT, connectivity preservation