Traffic Scheduling and Trajectory Optimization of RIS-assisted UAV-based IoT Network

Abstract

Unmanned Aerial Vehicles (UAV) have been very effective for data collection from widely spread Internet of Things Devices (IoTDs). However, in case of obstacles, the Line of Sight (LoS) link between the UAV and IoTDs will be blocked, resulting in severe deterioration of the available quality of service (QoS). To address this issue, the Recnfigurable Intelligent Surface (RIS) has been very effective, especially in urban areas, to extend communication beyond the obstacles, thus enabling efficient data transfer in situations where the LoS link does not exist. RIS is a cheap metasurface technology, which can extend cellular coverage in such circumstances. In this work, the goal is to jointly optimize the trajectory and thus minimize the energy consumption of UAVs on one hand and maximize the throughput, data rate, and energy efficiency of IoTDs on the other hand. As it is a mixed integer non-convex problem, Reinforcement Learning (RL); a class of Deep Learning (DL), is used which has proven to be more effective as compared to the conventional techniques e.g Successive Convex Approximation (SCA). In this work, three discrete RL agents i.e. Double Deep Quality Network (DDQN), Proximal Policy Optimization (PPO), and Proximal Policy Optimization (PPO) agent with Recurrent Neural Network (RNN) are tested with multiple RISs to enhance the data transfer between UAV and IoTDs. The results show that DDQN with multiple RIS provides better average throughput and Energy Efficiency (EE), while a single RIS system with the Proximal Policy Optimization (PPO) agent provides more average data rate and reduce the UAV’s fuel consumption when compared to other agents with multiple RIS..