Applied Network Calculus for Optimization of LoRaWAN Communication using Insights from Real-world Measurements of Energy Harvesting Performance
Authors
Mulvad, Thomas Gundgaard ; Frederiksen, Rasmus Sibbern
Term
4. semester
Education
Publication year
2024
Abstract
Inden for strukturel tilstandsovervågning (SHM) er der ofte behov for 'sæt-op-og-glem' trådløse sensorer, fordi vedligeholdelse ikke er mulig. Sådanne sensorer skal kunne køre uendeligt ved hjælp af energihøstning (EH). Udfordringen er at matche variationerne i den høstede energi med pålidelig trådløs dataoverførsel, fordi gennemsnitlige tal skjuler de tidsmæssige udsving, der påvirker driften. Denne afhandling præsenterer en metode til at anslå den maksimale pålidelige gennemstrømning (hvor meget data der kan leveres stabilt over tid) for et trådløst system med en given energikilde, i dette tilfælde en solcelle. Metoden bygger på Deterministic Network Calculus (DNC), en matematisk ramme til netværks- og ydeevneanalyse. Metoden valideres med virkelige målinger af både kommunikationspålidelighed og energihøstning ved hjælp af en implementeret EH-drevet sensorknude. Noden bruges i en accelereret målekampagne og emulerer flere LoRaWAN-enheder for at teste flere kommunikationsparametre samtidig. Målingerne viser, at en bestemt LoRaWAN-konfiguration giver den bedste balance mellem en fast pålidelighed og energiforbrug. De viser også, at længere pakker ikke nødvendigvis er bedre, fordi de er mere udsatte for kollisioner i ISM-båndet. Ved at modellere energiindstrømning og energiforbrug som flows anvendes netværkskalkyle til at bestemme den maksimale gennemstrømning.
In Structural Health Monitoring (SHM), maintenance is often not possible, so 'fit-and-forget' wireless sensors are needed. These sensors must run indefinitely using Energy Harvesting (EH). The challenge is to align time-varying harvested energy with reliable wireless transmissions, because simple averages hide timing effects that matter for operation. This thesis presents a method to approximate the maximum reliable throughput (how much data can be delivered consistently over time) of a wireless system powered by a given energy source, here a solar cell. The method is based on Deterministic Network Calculus (DNC), a mathematical framework for network performance analysis. The approach is validated with real-world measurements of both communication reliability and harvested energy using a prototype EH-powered sensor node. The node supports an accelerated measurement campaign and emulates multiple LoRaWAN devices to test several communication parameters at the same time. The results show that a specific LoRaWAN configuration offers the best balance between a fixed reliability target and energy consumption. They also show that longer packets are not always better, as they are more prone to collisions in the ISM band. By modeling energy inflow and consumption as flows, the network calculus analysis identifies the maximum throughput achievable under these constraints.
[This summary has been rewritten with the help of AI based on the project's original abstract]
Documents