A Deep Deterministic Policy Gradient Method for Optimizing Task Completion Time and Energy Efficiency in UAV-Assisted IoT Networks

Unmanned Aerial Vehicles (UAVs) have become increasingly pivotal in various Internet of Things (IoT) implementations due to their dynamic mobility and adaptability. This article examines the deployment of a rotary-wing UAV to efficiently manage data retrieval from dispersed IoT nodes within a UAV-assisted IoT framework. To support energy-efficient data collection, we incorporate uplink Non-Orthogonal Multiple Access (NOMA), enabling multiple IoT nodes to transmit data simultaneously. The core objectives of our research include improving the UAV navigational path, reducing energy consumption, reducing mission duration, and improving overall data collection efficiency. However, these goals present considerable difficulties, as conventional optimization techniques often fall short when faced with unpredictable time slots, resulting in a vast range of decision variables and the complexity of non-convex functions. We approach these challenges by framing the optimization tasks within a Markov Decision Process (MDP), employing the Deep Deterministic Policy Gradient (DDPG) algorithm to develop adaptive, robust UAV control strategies. This methodology facilitates dynamic adjustment of UAV operations, ensuring effective data collection and management. Numerical analysis validates the effectiveness of our method, showcasing significant improvements over existing relevant works.