Mathematical Modelling and Prediction of Interference Power in In-robot Subnetworks
Authors
Menholt, Lukas ; Friis, Lars Vedel
Term
4. semester
Education
Publication year
2022
Pages
92
Abstract
Future 6G “in‑X” subnetworks—small wireless networks embedded inside systems such as factory robots—must deliver very low delay, high data rates, and high reliability. This thesis studies the interference power (the strength of unwanted radio signals) produced by in‑robot subnetworks in a factory environment, aiming to analyze and predict how it evolves over time. We derive analytical expressions for the average interference power and for its auto‑correlation function (ACF), which measures how current interference relates to past values. These expressions involve high‑dimensional integrals, evaluated using Monte Carlo integration (random sampling). The analytical mean agrees with the mean obtained from simulations. For the ACF, the Monte Carlo estimate tends to overestimate the magnitude compared with simulations; however, it has the same shape, and an appropriate scaling aligns the two. We then model the interference power as a time series using an autoregressive model of order 20 (AR(20)). The AR(20) predictor can forecast the interference up to 8 ms ahead when the velocity is 2 m/s, and it outperforms a simple last‑value predictor across all tested settings.
Fremtidens 6G “in‑X” delnet – små trådløse net indbygget i fx fabriksrobotter – skal levere meget lav forsinkelse, høj datahastighed og høj pålidelighed. Denne afhandling undersøger interferens (styrken af uønskede radiosignaler) fra in‑robot‑delnet i et fabriksmiljø, med fokus på at analysere og forudsige dens tidslige adfærd. Vi udleder analytiske udtryk for den gennemsnitlige interferens og for dens autokorrelationsfunktion (ACF), som beskriver, hvor meget nutidens interferens hænger sammen med fortidens værdier. Udtrykkene indeholder højdimensionale integraler, som vi beregner ved hjælp af Monte Carlo-integration (tilfældig sampling). Det analytiske gennemsnit stemmer overens med gennemsnittet fra simulationer. For ACF’en har Monte Carlo-estimatet tendens til at overvurdere størrelsen i forhold til simulationerne; formen er dog den samme, og en passende skalering kan få dem til at falde sammen. Dernæst modellerer vi interferensen som en tidsserie med en autoregressiv model af orden 20 (AR(20)). AR(20)-prædiktoren kan forudsige interferensen op til 8 ms frem ved en hastighed på 2 m/s og overgår en simpel sidste-værdi-prædiktor i alle afprøvede indstillinger.
[This apstract has been rewritten with the help of AI based on the project's original abstract]
Keywords