Real-Time Digital Twin Platform: A Case Study on Core Network Selection in Aeronautical Ad-Hoc Networks
Lal Verda Cakir, Mihriban Kocak, Mehmet Özdem, Berk Canberk
TL;DR
This work tackles the lack of open-source, real-time digital twin platforms by delivering a Real-Time Digital Twin Platform tailored for aeronautical ad-hoc networks (AANETs). It combines a microservice architecture, a robust time-series data pipeline, and an interactive Flask/Grafana UI to enable real-time core network selection using data from the OpenSky Network and stored in InfluxDB, with Kapacitor handling streaming decisions. Key contributions include a robust ETL workflow with a projection-based data correction to handle missing data, Bayesian Information Criterion for cluster count selection, and KMeans clustering to drive network recommendations, all validated with real-time data achieving decision latencies under 3 seconds. The platform demonstrates practical value for real-time DT validation in aviation connectivity, offering an open-source framework that can adapt to dynamic data and potential data retrieval failures.
Abstract
The development of Digital Twins (DTs) is hindered by a lack of specialized, open-source solutions that can meet the demands of dynamic applications. This has caused state-of-the-art DT applications to be validated using offline data. However, this approach falls short of integrating real-time data, which is one of the most important characteristics of DTs. This can limit the validating effectiveness of DT applications in cases such as aeronautical ad-hoc networks (AANETs). Considering this, we develop a Real-Time Digital Twin Platform and implement core network selection in AANETs as a case study. In this, we implement microservice-based architecture and design a robust data pipeline. Additionally, we develop an interactive user interface using open-source tools. Using this, the platform supports real-time decision-making in the presence of data retrieval failures.