AIU1 – Applications, IoT, Use cases
Wednesday, 3 June 2026, 8:30-10:00, room Sala 6 (1st floor)
Session Chair: Foteini Setaki (Hellenic Telecommunications Organization, GR)
QoS-Aware Routing in Software-Defined Networking-Enabled Internet of Things via LSTM Learning-Based Optimization Algorithm
Seyed Salar Sefati (National University of Science and Technology Politehnica Bucharest, Romania); Atanas Georgiev Vlahov (Technical University of Sofia & Research and Development and Innovation Consortium, Bulgaria); Octavian Fratu and Simona V. Halunga (University Politehnica of Bucharest, Romania)
One of the main challenges in today’s SDN-enabled IoT networks is ensuring that Quality of Service (QoS) is managed proactively. Because of dynamic traffic patterns, timevarying link conditions, and the reactive nature of routing protocols, network control entities often fail to prevent congestion before network performance degrades. In this paper, we propose a QoS-aware routing framework that can proactively minimize link congestion and reliability loss. The proposed framework integrates Long Short-Term Memory (LSTM) learning with a Modified Black- Winged Kite Optimizer (M-BWKO). The LSTM model forecasts future path delay and reliability using historical telemetry data, while the optimizer selects the optimal routing configurations that minimize delay violations, reliability loss, and link congestion. Simulation results demonstrate that our predictive approach improves end-to-end delay, reliability, and network goodput compared with conventional and learning-only routing approaches, particularly under high traffic loads.
Building the Future of Rail: A 5G Edge Reference Architecture for Intelligent Monitoring in Germany
Sanket Partani (RPTU University Kaiserslautern-Landau, Germany); Anjie Qiu (RPTU Kaiserslautern-Landau, Germany); Marvin Reski and Andreas Weinand (RPTU University Kaiserslautern-Landau, Germany); Hans D. Schotten (RPTU Kaiserslautern-Landau, Germany)
The modernisation of Germany’s railway infrastructure demands innovative solutions that combine intelligence, scalability, and resilience against both technical failures and external disruptions. Centralised monitoring systems, though effective in data aggregation, face inherent challenges such as latency, bandwidth limitations, and dependence on continuous connectivity. To address these issues, this work presents a 5G edge reference architecture designed to enable large-scale, intelligent, and secure railway infrastructure monitoring across the German rail network. Based on the density of the sensors, the proposed architecture integrates distributed edge and far-edge computing with AI-based analytics and secure 5G communication. In this configuration, sensors capture mechanical and environmental signals and transmit the data to the nearest edge node, where AI models analyse the signals to detect anomalies, wear, and potential sabotage in real time. This decentralised architecture enhances system resilience, energy efficiency, and data security while supporting low-latency, safety-critical applications. As part of the concept development, an extensive analysis of current railway switch configurations across Germany is conducted to inform the design of both decentralised and centralised architectures. This comparative study enables the evaluation of energy costs, data transfer requirements, and system performance under different deployment models. The findings serve as the foundation for the proposed decentralized, 5G edge architecture, which is intended as a reference model for future railway monitoring systems. A preliminary measurement campaign within a 5G campus network is carried out to emulate the real-world network conditions and validate the architecture’s key principles. While further validation is ongoing, initial results confirm the feasibility and scalability of the proposed approach, marking an important step toward a resilient, intelligent, and future-ready railway infrastructure in Germany.
A Case Study: QoS Management in Mobile Communications with the Asset Administration Shell
Mike Reichardt (Deutsches Forschungszentrum Für Kuenstliche Intelligenz GmbH, Germany); Daniel Buch (Deutsches Forschungszentrum Fuer Kuenstliche Intelligenz GmbH, Germany); Hans D. Schotten (RPTU Kaiserslautern-Landau, Germany)
Digitalization is becoming increasingly essential in modern industrial environments, driving the development of digital representations of assets for purposes such as simulation, monitoring, and lifecycle management. These digital representations are not restricted to physical components but can also describe virtual assets, processes or entire systems, including communication infrastructures. Building on this concept, 5G-ACIA has proposed an approach for managing industrial 5G systems using digital twins realized through the Asset Administration Shell (AAS). In this approach, they suggested to use two different AAS models a 5G network AAS for the network and a 5G UE AAS for each user equipment (UE). Prior research has demonstrated the potential of AAS-driven management of industrial 5G deployments. However, existing studies have been limited to small-scale scenarios involving no more than five UEs. In this work, we present a quantitative analysis of the scalability and performance of an AAS-based management approach for industrial 5G and upcoming 6G systems with 40 UEs. The evaluation is conducted using Amarisoft for the emulation and shows performance gains between 13% to 48% in different metrics.
DoA Estimation-Driven Environmental Awareness: Inverse Ray Tracing Approach
Jintong An and Selma Saidi (TU Braunschweig, Germany)
In this paper, a novel scene reconstruction approach-inverse ray tracing (iRT)-is proposed to support environmental awareness with low computational complexity and high robustness for IoT applications: By exploring the implicit associations of the scene geometry with the arrival direction of reflection signals, the corresponding reflection plane can be reconstructed efficiently during iterations. Subsequently, Hough voting strategy is adopted, which is applied for accurate and robust plane reconstruction with low computational cost, making it suitable for embedded Internet of Things (IoT) implementations. Through validation in indoor scenario our proposed iRT process is validated to achieve plane reconstruction of high accuracy and flexibility.
Advancing NB-IoT NTN Readiness: Experimental Demonstration of a Smart Mobility Pilot in the Pyrenees
Marti Fernandez Pons (Fundació i2CAT, Internet i Innovació Digital a Catalunya, Spain); Marcel Marin-de-Yzaguirre (i2CAT Foundation, Spain); Alex Sells (Fundació I2CAT, Internet i Innovació Digital a Catalunya, Spain); Oriol Fusté (Universitat Politècnica de Catalunya, Spain & i2CAT Foundation, Spain); Francesc Betorz (i2CAT Foundation, Spain); Ramon Ferrús and Marco Guadalupi (Sateliot, Spain); Joan Francesc Muñoz-Martin (i2CAT Foundation, Spain)
The lack of terrestrial network coverage in remote areas poses a significant barrier to implementing data-driven management for nature environment preservation. This work presents the experimental demonstration of a Smart Mobility pilot deployed in the Vall de Boí (Spanish Pyrenees), which integrates an autonomous Edge-AI counting system with the Minairó LEO satellite mission. The system leverages the recent 3GPP Non-Terrestrial Networks standardization to enable direct-to-satellite Narrowband IoT connectivity. We present the complete end-to-end functionality of the solution, tracing the data flow from autonomous acquisition at the edge to the final user interface. Furthermore, this work presents a validation study of the Minairó system, confirming its suitability as an enabling technology for such remote use cases while simultaneously characterizing its actual operational performance in in-field conditions.
