Reconfigurable intelligent surface (RIS) assisted throughput maximization via deep reinforcement learning in UAV-powered IoT networks

This master thesis deals with maximizing throughput using Reconfigurable Intelligent Surfaces (RIS) through Deep Reinforcement Learning in Unmanned Aerial Vehicle (UAV)-powered Internet of Things (IoT) networks. The main purpose of this study is to ensure that a UAV collecting data from IoT devices positioned in a region can collect maximum data from these devices, as well as increase energy efficiency and land safely in the target area. For this purpose, a UAV was deployed to collect data from various IoT devices more effectively and this UAV was trained for different environmental conditions. The study consists of two phases. In phase 1, an algorithm was developed to increase the directional capacity and reconnaissance capability of the UAV. In phase 2, the RIS was integrated into the system and the effect of the RIS was analysed. RIS technology is used to improve the reliability, safety, capacity and coverage of communication between IoT devices and UAV. In particular, when there is no direct line of sight between an IoT device and the UAV or when there is an obstacle (building, tree, etc.) between them, RIS can help to better transmit the signals from an IoT device to the UAV. In this study, we show that the use of RIS technology improves the communication between IoT devices and the UAV and increases the amount of data collected while maintaining energy efficiency. The simulation results show that increasing the directional capacity and reconnaissance capability of the UAV increases the data collection performance by 8,18% and the amount of data it can collect per unit of energy by 6,91%. Leveraging RIS technology also improves the performance of data collection by 10,7% and the performance of collected data per unit energy by 22,64%. In addition, the UAV, which was trained with the Double Deep Q-Network (DDQN) algorithm, one of the deep reinforcement learning algorithms, determined the most suitable route for collecting data from IoT devices and was able to land successfully. As a result, this study revealed that the use of RIS technology in UAV-powered IoT networks has great potential in terms of improved data collection and energy efficiency. The results of this study can be used for the development and optimisation of future UAV-powered IoT applications.