Decentralized Multi-Hop Federated Reinforcement Learning for Energy-Efficient and Secure Routing in LoRaWAN-Based Smart City Infrastructure

LoRaWAN (Long Range Wide Area Network) has emerged as a foundational technology in smart city infrastructures due to its capability to facilitate long-range, low-power communication between Internet of Things (IoT) devices. However, the inherently resource-constrained nature of these devices, coupled with dynamic network conditions, necessitates the development of advanced routing strategies that balance energy efficiency, data security, and scalability. This paper introduces a novel hybrid framework that integrates Federated Learning (FL) and Reinforcement Learning (RL) to enable decentralized, multi-hop routing in LoRaWAN-enabled smart cities. By leveraging localized model training (FL) and adaptive decision-making (RL), the framework addresses core challenges such as node mobility, trustworthiness, and network longevity. Through simulation-based evaluation, the proposed system demonstrates significant improvements in energy conservation, routing robustness, and protection against eavesdropping and tampering attacks, offering a scalable solution for future smart urban ecosystems.