Multi-Objective Reinforcement Learning for Data Collection-Energy Consumption Trade-off in UAV-Aided IoT Networks

Unmanned Aerial Vehicles (UAVs) are increasingly pivotal for data harvesting in the context of IoT networks, leveraging their exceptional mobility and flexibility. Traditional and AI-driven approaches for UAV data collection typically focus on optimizing a single or a fixed combination of objectives, such as data quantity, energy efficiency, or data freshness. However, these methods often fall short in balancing the diverse tradeoffs in complex, real-world tasks with conflicting objectives. Our work introduces a novel extension of the Discrete Soft Actor-Critic algorithm to the multi-objective framework. This approach enables UAVs to learn and adapt to diverse tradeoffs between data collection and energy consumption. Simulation results demonstrate significant improvements in stability and performance over existing deep multi-objective reinforcement learning algorithms.