Unsupervised Learning Clustering and Dynamic Transmission Scheduling for Efficient Dense LoRaWAN Networks

Mohammed Alenezi,Shihab Jimaa,Atm S Alam,Kok Keong Chai,Yue Chen

doi:10.1109/access.2020.3031974

Abstract

Long-Range (LoRa) communication technology is considered as a promising connectivity solutions for Internet of Things (IoT) dense applications. In particular, LoRa has drawn the interest due to its low power consumption and wide area coverage. Despite the benefits of LoRaWAN protocol, it still suffers from excessive random and simultaneous transmissions due to the adoption of ALOHA protocol. Therefore, resulting in severe packet collision rate as the network scales up. This leads to continuous retransmission attempts, which in return increase the transmission delay and energy consumption. Thus, this paper proposes a dynamic transmission Priority Scheduling Technique (PST) based on the unsupervised learning clustering algorithm to reduce the packet collision rate and enhance the network's transmission delay and energy consumption. Particularly, the LoRa gateway classifies the nodes into different transmission priority clusters. While the dynamic PST allows the gateway to configure the transmission intervals for the nodes according to the transmission priorities of the corresponding clusters. This work allows scaling up the network density while maintaining low packet collision rate and significantly enhances the transmission delay & the energy consumption. Simulation results show that the proposed work outperforms the typical LoRaWAN and recent clustering & scheduling schemes. Therefore, the proposed work is well suited for dense applications in LoRaWAN.

Highlights

T HE Low-Power Wide Area Networks (LPWAN) technologies have been increasingly researched and deployed as a promising solution for serving Internet of Things (IoT) applications
Given the unlabeled data delivered by the nodes, the limited resources for the devices and the random transmission behaviour of LoRaWAN due to adapting ALOHA protocol, this paper introduces the use of unsupervised learning clustering algorithm (K-Means) as a base for the dynamic transmission Priority Scheduling Technique (PST)
The use of the unsupervised clustering algorithm K-Means in LoRaWAN network has shown a great impact in reducing the collision rate and higher Packet Delivery Rate (PDR), while maintaining low energy consumption and transmission delay

Summary

Introduction

T HE Low-Power Wide Area Networks (LPWAN) technologies have been increasingly researched and deployed as a promising solution for serving Internet of Things (IoT) applications. Long-Range (LoRa) technology via its LoRaWAN protocol [1], has shown a very attractive platform due to its low energy consumption and wide area coverage. One main drawback associated with LoRaWAN is the vulnerability to high packet collision rate. This is due to the adaption of ALOHA communication protocol, where LoRa nodes initiate packet transmissions without the presence of Listen Before Talk (LBT) protocol [2], [3]. In order to compensate for the absence of LBT protocol, LoRaWAN provides different Spreading Factors (SF) based on the LoRa physical layer Chirp Spread Spectrum (CSS) technique to allow simultaneous packet transmissions. Adapting LoRaWAN to serve dense applications remains an open challenge

Objectives

Findings

Discussion

Conclusion