Distributed Learning for Low Latency Machine Type Communication in a Massive Internet of Things

TL;DR

This paper introduces a finite memory multi-state sequential learning framework for IoT devices to efficiently share communication resources, prioritizing urgent messages and reducing delays in massive IoT networks.

Contribution

It presents a novel self-organizing learning method that accounts for IoT device limitations, enabling dynamic resource reallocation for critical messages.

Findings

Effective delay reduction for critical messages.

High success rate in resource learning across network conditions.

Framework converges and adapts to various IoT scenarios.

Abstract

The Internet of Things (IoT) will encompass a massive number of machine type devices that must wirelessly transmit, in near real-time, a diverse set of messages sensed from their environment. Designing resource allocation schemes to support such coexistent, heterogeneous communication is hence a key IoT challenge. In particular, there is a need for self-organizing resource allocation solutions that can account for unique IoT features, such as massive scale and stringent resource constraints. In this paper, a novel finite memory multi-state sequential learning framework is proposed to enable diverse IoT devices to share limited communication resources, while transmitting both delay-tolerant, periodic messages and urgent, critical messages. The proposed learning framework enables the IoT devices to learn the number of critical messages and to reallocate the communication resources for the periodic messages to be used for the critical messages. Furthermore, the proposed learning framework explicitly accounts for IoT device limitations in terms of memory and computational capabilities. The convergence of the proposed learning framework is proved, and the lowest expected delay that the IoT devices can achieve using this learning framework is derived. Furthermore, the effectiveness of the proposed learning algorithm in IoT networks with different delay targets, network densities, probabilities of detection, and memory sizes is analyzed in terms of the probability of a successful random access request and percentage of devices that learned correctly. The results show that the proposed learning algorithm is very effective in reducing the delay of urgent, critical messages by intelligently reallocating the communication resources allocated to the delay-tolerant, periodic messages.