An investigation into the application of machine learning algorithms in LoRaWAN towards the improvement of network performance

Abstract

LoRaWAN (Long Range Wide Area Network) has emerged as the de facto communication protocol for smart cities application. The number of IoT devices deployed is expected to reach 8 billion by end of this decade for which LoRaWAN networks are expected to be highly scalable, flexible, provide highest Quality of Service and data security. When wireless networks are scaled without improving the communication systems infrastructure, performance of the network decreases. Improving the hardware incurs huge cost investment and cannot be deployed rapidly, therefore it is important to identify potential points in a network to effectively scale a network with minimal hardware changes and maximize the scalability using algorithms at the network server. Informed decisions regarding the network performance and scalability are performed in the network server. To improve the performance of a network it is first necessary to understand the LoRa protocol. To understand the LoRa protocol it is necessary to understand both the physical and Medium Access Layer of the protocol. A real time implementation of LoRaWAN protocol was implemented in a dense urban environment. The LoRaWAN network was implemented in Coimbatore city, TamilNadu, India. The nodes were placed in different locations and the data set is created. The performance analysis of LoRaWAN are plotted and interpreted. newline