Tuesday, December 7, 2021
Reliability and capacity of low power and long range networks
This thesis focuses on Low Power Wide Area Networks (LPWAN), a new class of wireless networks which aim to provide massive connectivity at a low cost in the context of the Internet of Things (IoT). This thesis focuses more particularly on the LoRaWAN technology, which is an open protocol based on the LoRa proprietary modulation. LoRaWAN knows a significant success and is now established as one of the major players in LPWAN networks. 
A typical LoRaWAN network is composed of a massive number of end devices (ED), i.e. connected objects, a set of gateways (GW) acting as bridges between the LoRa network and the Internet Protocol (IP) network and finally a network server (NS). EDs transmit their data via LoRa modulation with an random channel access method (unslotted Aloha-like), i.e. without any form of coordination. The GWs transmit to the NS via the IP network all the data captured on their LoRa interface. The NS is responsible for processing the received data. LoRaWAN is an uplink oriented network, meaning that the overwhelming majority of the traffic is expected to be from the EDs to the NS. However, LoRaWAN also allows rare downlink transmissions, from the NS to the EDs. 
Due to the massive number of EDs, the low transmission powers and the random channel access method, data is sometimes lost by the LoRa link. 
This thesis proposes solutions to make LoRaWAN communication more reliable, i.e. to deliver more than 99% of the application data. The reliability gain must be achieved in a realistic manner with respect to the capacity of the network, i.e. the number of EDs that can be served by the network in the same area. The consequence of this constraint is twofold: on the one hand the Time On Air (TOA) for the uplink transmissions must be kept as low as possible in order to avoid network congestion as much as possible, and on the other hand the TOA for downlink transmissions must be kept extremely low due to the limited downstream capability of LoRaWAN. 
With the goal of a reliable LoRaWAN communication, this thesis firstly proposes an in-depth study of the characteristics of the LoRaWAN link, based on experimental measurements in a public urban network. Our characterization of the LoRaWAN link leads us to conclude that the reliability mechanisms currently existing in the LoRaWAN protocol are unfit to provide a highly reliable link. We therefore propose adaptations to make the LoRaWAN protocol reliable while conserving the technology scalability. 
The thesis proposes two types of adaptations: 
The first adaptation consists in adding an error recovery protocol overlay, transparent for LoRaWAN, based on error correcting codes applied transversely to the packet flow. This approach makes it possible to reconstitute all of the data transmitted despite packet losses. The thesis proposes two distinct algorithms for this error recovery aspect, one based on the Reed-Solomon correcting code, and the other based on a correcting code derived from Low Density Parity Check (LDPC) codes. This thesis evaluates the performance of the two proposed error recovery algorithms. 
The second proposed adaptation consists in reviewing the distribution of the transmission parameters of the EDs in the network. It is based both on the characterization of the LoRaWAN channel and on the observation that the proposed error recovery overlay makes it possible to obtain high reliability while making it possible to be tolerant to a certain threshold of packet loss. This thesis therefore proposes to review the LoRaWAN's Adaptive Data Rate (ADR) algorithm. The ADR is an algorithm which supports over-the-air and dynamic configuration of the EDs transmission parameters. We propose in this thesis, to optimize the ADR in order to reduce the TOA of the EDs to the minimum required.
Mis à jour le 3 December 2021