Approved
Expanding a LoRaWAN network for cost efficiency improvement
Eva Jurado (HT-17)
Start
2017-12-01
Presentation
2018-11-08
Location:
Finished:
2018-11-15
Master's thesis:
Abstract
This thesis will delve into network architecture planning when running a larger LoRaWAN network. LoRAWAN prides itself as an LPWAN meaning its designed to consume very little power, thus reducing the maintenance cost of sensors. However, when deploying a gateway it will naturally have a varying distance to every sensor, and a sensor far away will need to send information in a way that consumes more of it's battery power compared to a sensor closer to the gateway. In LoRaWAN this mechanism uses a feature called Adaptive Data Rate which can reconfigure sensors to more or less encoding bits when sending. A message containing the same information can take 40ms or 1.4 seconds to transmit. Since battery replacement is costly, especially inside apartments which have significant administration overhead in the range of 600 SEK/entrance, it might be cost effective to put up gateways inside buildings to increase the battery life of sensors by increasing the link-budget and reducing transmission times. Especially as the number of sensors is expected to increase over time this is a dynamic and constantly changing calculation. Furthermore, LoRaWAN introduces roaming in their latest specification which puts additional options for the sensor and/or network owner. How would one design a network for optimizing maintenance cost and minimize (expensive) roaming to other networks or decide to employ more gateways as a sensor or network owner? Other parameters to consider is the gateway load from sensors which decreases with shorter message lengths due to increased link budgets. The thesis will consist of both a theoretical and practical part. The theoretical part will be to study relevant material in statistic and network architecture. The practical part will include physical measurements in the city of Lund, inside buildings and surrounding environments to develop models for different scenarios and verify them.
Supervisor: Anders Hedberg (Sensefarm AB) and Björn Landfeldt (EIT)
Examiner: Christian Nyberg (EIT)