GNU Radio implementation for Multiuser Multi-Armed Bandit learning algorithms in IoT networks

Abstract

Novel access schemes based on multi-armed bandit (MAB) learning approaches has been proposed to support the increasing number of devices in IoT networks. In the present work, a GNU radio framework is implemented to recreate an IoT network where IoT devices embedding MAB algorithms are able to learn the availability of the channel for their packet transmissions to the gateway. It allows to incorporate several IoT users recognized by an identifier (ID), and provides a gateway to handle a large number of IDs as well as the packet collisions among IoT devices. The experimental results show that the introduction of learning mechanism in access schemes can improve the performance of the network.