Reinforcement Learning-Based Task Scheduling for IoT Applications in Long-Range Wide Area Networks

To address the challenges of effective resource allocation in low-power wide-area networks, this thesis examines the scheduling of end devices in Internet of Things (IoT) applications using LoRaWAN technology. The primary objective of this research is to utilize reinforcement learning (RL) to enhance quality of service (QoS) metrics, including energy efficiency, throughput, latency, and reliability. This objective was achieved through a simulation-based approach that assessed the performance of the RL-based scheduling algorithm using NS-3 simulations. The key findings indicate that, compared to existing scheduling methods, the RL agent significantly enhances data transmission reliability and increases network throughput. Additionally, the proposed approach effectively reduces average system latency and overall energy consumption, leading to improved network resource utilization. These results suggest that applying RL to task scheduling in LoRaWAN networks can provide a scalable and reliable solution to existing challenges, ultimately contributing to more intelligent and sustainable IoT systems. Overall, this study concludes that RL-based techniques can enhance resource management in dynamic and resource-constrained environments.