Energy Efficient Reinforcement Learning Based Adaptive Resource Allocation for LoRa Networks

In the last few years, a surge in IoT applications has ramped up the need for effective and dependable data transmission in LoRa-based systems. Yet, traditional resource allocation methods in LoRa systems face major drawbacks such as higher packet loss, interference, excessive energy use, limited coverage, slow transmission speeds, and increased operational expenses. To tackle these issues, this study introduces a new hybrid optimisation framework that combines Hybrid Reinforcement Learning, named as Double Deep Q-Learning based Actor-Critic mechanism (Hy-DeoQ-AC), with a hybrid Levy Flight Assisted Rabbit optimisation algorithm (Hy-LevRBO). Hy-DeoQ-AC mechanism learns optimal network configurations dynamically by engaging with the environment, concentrating on key transmission parameters like spreading factor, transmission power, and channel selection to satisfy strict Quality of Service (QoS) requirements of IoT devices. Additionally, the hybrid optimisation gains from Hy-LevRBO, which fine-tunes chosen parameters and boosts capability to evade local optima. Thus, this combined strategy greatly enhances energy efficiency, maximises throughput, extends transmission range, and reduces latency in LoRa networks. The Comprehensive experimental analysis attains a throughput of 56.8471(bits/s), energy efficiency of 16.1364 (bits/J), which confirms the proposed model's superiority and achieves better performance across various metrics. This research offers an energy-efficient solution for IoT communications.