OPTIMIZING IOT NETWORKS : A GNU RADIO IMPLEMENTATION OF MULTI ARMED BANDITS LEARNING

Abstract

To monitor large-scale systems like smart grids and smart cities effectively, dedicated networks for Internet-of-Things (IoT) applications are being developed. These networks, such as LoRaWAN and SigFox, utilize Low Power Wide Area Networks in unlicensed frequency bands. LPWANs handle a significant number of devices that transmit only a few packets per day or week. To optimize energy consumption in end devices, these networks employ ALOHA-based Medium Access Control protocols. An important challenge in designing MAC solutions for IoT is to enhance network performance and reduce the Packet Loss Ratio while maintaining long battery life for end devices. Since many IoT standards operate in unlicensed bands, it is crucial to find solutions that minimize PLR caused by interference from other standards and networks sharing the same frequency band without coordination. The interfering traffic originates from different standards and networks, making it uncontrollable and unevenly distributed across channels. Recently, Multi-Armed Bandit algorithms have emerged as a potential solution for improving IoT network performance, particularly in LPWAN settings.