Low Power Wide Area Networks (LPWAN) are used in many Internet of Things (IoT) applications. Long Range Wide Area Network (LoRaWAN) is a well-known LPWAN protocol, offering wide coverage with low power consumption and a low data rate.
LoRaWAN is a data link layer protocol developed by the LoRa Alliance to provide long-range connectivity with low power consumption. It relies on the LoRa long-range physical layer developed by Semtech, based on Chirp Spread Spectrum (CSS) modulation.
The default LoRaWAN network topology is a star-of-stars, where end devices transmit directly to gateways. The network consists of end nodes, gateways, and a LoRaWAN network server. The gateway receives packets from the end nodes and forwards them to the network server, which manages the network, including end-node address checking, acknowledgements, frame counters, and so on.
Many studies have shown that a LoRaWAN network can reach coverage of several kilometres in open areas and rural environments. However, performance drops in areas with obstacles such as buildings or mountains. Dense networks with many users also see degraded performance due to interference and packet collisions [1].
The LoRa Alliance has therefore recently proposed specifications introducing relay nodes into LoRaWAN networks [2]. Several works in the literature have also studied this evolution [3][4]
This topic is part of the ANR MORAL project, led jointly by I3S, LIG and IRIT, which aims to design models and algorithms for LoRaWAN relaying that is both high-performing and energy-efficient, relying in particular on the LIG's WalT/WalToR/FeniX experimental platform.
- C. Caillouet, M. Heusse, F. Rousseau, Optimal SF Allocation in LoRaWAN Considering Physical Capture and Imperfect Orthogonality, IEEE Globecom 2019.
- https://resources.lora-alliance.org/technical-specifications/ts011-1-0-1-relay
- E. Lumet, A. Le Floch, R. Kacimi, M. Lihoreau, A.L. Beylot, LoRaWAN relaying: Push the cell boundaries. ACM MSWiM 2021.
- J.R. Cotrim, J.H. Kleinschmidt, An analytical model for multihop LoRaWAN networks. Elsevier Internet of Things, Volume 22, 100807, July 2023.
The goal of this PhD is to optimize the use of relay nodes in LoRaWAN networks and their impact on network capacity, through a sound mathematical model integrating:
- the characteristics defined in the LoRa Alliance specifications for the use of these relays,
- analytical physical models of propagation and interference in these networks.
The questions the PhD candidate will address include:
- What is the maximum achievable capacity for a given number of relays?
- What is the impact of adding relays to the network on its capacity?
- What is the optimal placement of relays?
- How can channel, SF, power, and relay resources be jointly optimally allocated?
The results will aim to assess the number of relays needed depending on the number of network users, their locations, and the cost and energy gains of deploying them.
Methodologically, the thesis will rely on operations research tools: formulating the dimensioning, relay placement and resource allocation problems as mixed-integer linear programs (MILP), solving them with decomposition techniques (Lagrangian relaxation, Benders decomposition, column generation), and then exploring reinforcement-learning approaches to approximate optimal solutions on larger-scale instances, benchmarked against the MILP solutions. Models and algorithms will be validated through simulation (ns-3) and then experimentally on the WalT/WalToR/FeniX platform developed at LIG.