Optimization Design in RIS-Assisted Integrated Satellite-UAV-Served 6G IoT: A Deep Reinforcement Learning Approach

Abstract

Satellite networks have been emerged as a critical part of the next-generation wireless networks. However, the high transmission latency, highly dynamic channel conditions and energy resource constraints of Internet of Things (IoT) devices pose a challenge to performance improvements. To tackle above issues, technologies such as integrated satellite-unmanned aerial vehicle-terrestrial networks (IS-UAV-TNs), deep reinforcement learning (DRL), reconfigurable intelligent surface (RIS) are highly anticipated in 6G IoT. In this article, we consider the application of RIS to IS-UAV-TNs to reshape wireless channels by controlling the phase shift of the scattering elements. The dynamic configuration of the RIS reflection unit poses a high-dimensional problem, making beamforming optimization challenging. We focus on discussing the optimization method of integrating DRL in RIS-assisted IS-UAV-TNs, which offers flexibility in scenarios where precise channel state information (CSI) is unknown. To illustrate the advantage of the DRL framework in RIS-assisted IS-UAV-TNs, we design a representative communication scenario, where the results are provided according to the considered scenario. Finally, potential future research directions and challenges are presented.