Deep Learning Collision Aware Transmission Scheduling for Dense LPWANs

Abstract

The rapid growth of the Internet-of-things (IoT) in recent years has led to the development of new access technologies targeting low-power wide area networks (LPWANs). LPWAN is a promising solution for long-range and low-power IoT and machine-to-machine (M2M) communication applications. However, practical LPWAN deployments suffer from intense collisions due to uncoordinated transmissions, given the dense deployment of devices and wide coverage area. Therefore, this paper proposes a deep learning-based collision aware transmission priority scheduling technique (CA-PST) to mitigate high packet collision rate in ultra-dense wireless networks. The proposed CA-PST is applied to the low-range wide area network (LoRaWAN) platform where the LoRa gateway configures the nodes to a particular transmission protocol classes based on the predicted number of packet collisions. Aided by unsupervised learning clustering, the nodes located at higher transmission priority clusters will be allowed to communicate with the gateway using class C and therefore avoid packet collisions. Whereas nodes in lower transmission priority clusters communicate with the gateway using class A. Therefore, the CA-PST offers a fair trade-off between the packet delivery rate (PDR) and total energy consumption. Extensive simulation results show that using CA-PST can significantly improve the LPWAN reliability in terms of PDR in ultra-dense networks.

Keywords

• Collision rate
• deep learning
• energy consumption
• IoT
• LPWAN
• packet delivery rate