Hibridni događaj

Program događaja

utorak, 26.5.2026 10:00 - 13:30, Galija, Hotel Admiral, Opatija 10:00 - 10:15 Otvaranje  Predsjedateljica: Željka Car  10:15 - 10:45 Pozvano predavanje Predsjedatelj:   • Pavle Skočir (Axiros, Zagreb) Primjena sustava za upravljanje i nadzor telekomunikacijskih uređaja u energetici  10:45 - 11:00 Odmor  11:00 - 12:15 Primjena umjetne inteligencije i inteligentni sustavi Predsjedatelj:   1. A. Rista, E. Xhaferra, L. Mukli (Aleksander Moisiu University of Durres, Durres, Albania) An Overview of LLM-Based AI Chatbots  Large Language Models (LLMs) have greatly advanced the field of AI chatbots, making multimodal generation, long-context reasoning, complex task solving, and automated workflows possible. This article examines the progression of chatbot technologies, from early rulebased and retrieval-oriented models to contemporary generative and multimodal LLM-based systems. Models are examined in terms of architectural design, performance evaluation, applications, limitations, strengths, safety, and ethical considerations, based on the analysis of technical reports, benchmark results, and analyses reported in the literature. From this perspective, a classification of them across domains has been done, highlighting that different chatbot models have different strengths and weaknesses across applications, and model selection should balance performance, security, and user requirements. The results indicate that autoregressive Transformer-based architecture demonstrates great performance in generative and opendomain chatbots, while encoder and task-oriented architecture models are suitable for specific Natural Language Processing (NLP) tasks. Integration of Mixtureof-Experts (MoE) approaches with multimodal and longcontext shows exceptional performance in long-context understanding, complex task solution, and automation workflow. This paper presents a valuable contribution, offering guidelines for users, developers, and researchers. 2. A. Rista, E. Xhaferra, E. Tata (Aleksander Moisiu University of Durres, Durres, Albania) Impact of Reward Design on Reinforcement Learning Performance for Autonomous Parking Agents  Autonomous parking is a challenging task requiring precise maneuvering and obstacle avoidance within constrained spaces. Deep Reinforcement Learning (DRL) offers a promising approach by allowing an agent to learn complex parking behaviors through trial and error. In this study, we investigate how the design of the reward function impacts the learning performance of a DRL-based self-parking agent. We develop a 3D simulated parking lot environment using Unity ML-Agents Toolkit and train an agent with a Proximal Policy Optimization (PPO) algorithm. The agent’s training progress and performance are analyzed over training episodes. Results show that an informative, well-shaped reward function greatly accelerates learning compared to a sparse reward, leading the agent to achieve successful parking in roughly 50% of attempts by the end of training (up from near 0% at the start). The learned policy not only improves the success rate but also parks more efficiently, requiring fewer steps and less time per maneuver. These findings highlight the critical impact of reward design on the effectiveness of reinforcement learning for autonomous parking. 3. D. Teliti, O. Shehu, J. Kevrić (International Burch University, Sarajevo, Bosnia and Herzegovina), B. Karlik (Istanbul Esenyurt University, Istanbul, Turkey) Machine Learning and Deep Learning Approaches for Breast Cancer Prediction: A Review  Breast cancer is still the most common cancer and leading cause of death among women all over the world; thus, there is an urgent need for reliable predictive models for early diagnosis, prognostic evaluation, and personalization of therapeutic approaches. In this review, we describe the existing computer-aided design (CAD) techniques on BC prediction and prognostication-based studies for the time period of 2017 through 2025 using ML and DL. We delineate methods along three axes: classical ML classifiers and deep learning architectures and techniques for ensembling models to leverage several types of classifiers through bagging, boosting, stacking, or weighted voting. We demo these strategies on benchmark datasets such as TCGA-BRCA, Wisconsin Breast Cancer, and METABRIC using accuracy, AUC-ROC, and survival analysis metrics with a focus on feature engineering pipelines like quantum-inspired transformation and particle swarm optimization for biomarker selection. Ensemble approaches by combining various classifiers using a highly refined feature set have the best performance to achieve the highest predictive accuracy, with more than 96% overall accuracy and AUC-ROCs surpassing 0.98, as was reported in recent publications. We discuss open challenges such as data standardization, model interpretability, and multi-institutional validation and suggest future research directions. 4. O. Shehu, D. Teliti, J. Kevrić (International Burch University, Sarajevo, Bosnia and Herzegovina), B. Karlik (Istanbul Esenuyrt University, Istanbul, Bosnia and Herzegovina) Deep Learning-Based Intrusion Detection for Healthcare IoT: A Systematic Review  The IoMT revolution has enabled real-time monitoring, remote diagnostics, and individualized therapy in modern healthcare through a growing array of IoMT devices. But the digital revolution has also expanded the attack surface for cyberthreats, and health care has become a popular target for bad actors. In this study, we review deep learning for intrusion detection in healthcare IoT networks. We survey twenty-five peer-reviewed articles published between 2017 and 2025 and classify approaches based on model structure, data utilization, and performance metrics. In our investigation, we discovered that RNNs, like bidirectional long short-term memory (BLSTM) and gated recurrent units (GRU), have more capability in modeling the temporal property of attacks through bigger IoT datasets with accuracy greater than 98%. CNNs work effectively in feature extraction from structured traffic data, and the hybrid models with mixed architectures are considered a potential solution for sophisticated threat scenes. We point to important research issues and limitations: synthetic datasets' overrepresentation, lack of real-world validation, and computational constraints in IoT devices are not addressed. The suggested study may be used as a solid foundation for researchers and developers in need of creating AI-based security tools in health care centers that are less under consideration within this review. 5. D. Berisha, E. Mulaj, J. Dobruna, M. Ibrani (University of Prishtina, Faculty of Electrical and Computer Engineering, Prishtine, Kosovo), M. Volk (University of Ljubljana, Faculty of Electrical Engineering,, Ljubljana, Slovenia) AI-Based Prediction of Low-Frequency Magnetic Field Exposure in Diverse Environments  Artificial intelligence (AI) -based prediction models were developed to estimate low-frequency magnetic field exposure levels using data from an in-situ measurement campaign conducted in Kosovo. Measurements of magnetic flux density (µT) were collected across four environments: indoor home, indoor office, outdoor, and transport, together with contextual variables including timestamp, geographic location, area type, people count, and environmental descriptors. These heterogeneous inputs were processed to train and evaluate machine-learning models designed to capture exposure variability. Model performance was assessed using standard error metrics, achieving a root mean square error of 0.0161 µT, mean absolute error of 0.0106 µT, coefficient of determination of 0.2517, and mean absolute percentage error of 4.28%. Although performance varied across environments, the low absolute error relative to measurement precision indicates functional capability of the system within the studied domain. The developed framework enables interpolation of exposure estimates between measurement points, identification of higher-exposure locations for targeted monitoring, and preliminary assessments of exposure levels in diverse environments. These findings indicate that AIdriven approaches can complement traditional measurement campaigns by enhancing spatial and temporal resolution of exposure data. The study contributes to advancement in exposure assessment and provides a foundation for scalable, context-aware prediction systems in environmental and human electromagnetic exposure research. 12:15 - 12:30 Odmor  12:30 - 13:30 Budući mrežni sustavi i distribuirana inteligencija Predsjedatelj:   6. D. Zorić, T. Grgić, B. Pavelić Grbić, D. Kljajić, T. Rihtarec, A. Grgurić (Ericsson Nikola Tesla, Zagreb, Croatia) QoE-driven Traffic Prioritization Model for XR Communication and Interaction Through a Dynamically Updated Digital Twin in Private 5G Networks  This paper proposes a traffic prioritization model for Extended Reality (XR) communication realized through a dynamically updated Digital Twin (DT) in private 5G networks. A target communication environment is described, consisting of a private 5G network as a central communication point (i.e., managed network segment) and the surrounding communication systems (i.e., unmanaged network segments). In this approach, the relationship between the 5G network QoS classes (5QI) and the Internet's DSCP QoS classes is investigated, making it possible to achieve endto-end QoE assurance across both managed and unmanaged network segments, by prioritizing more QoE-sensitive traffic. Various content-rich types of communication are considered in this work, which can usually be found in real-time interaction between dynamically updated digital twins and the real world, such as: XR traffic flows, signaling and control data exchange, splat data, audio/video communication and streaming, remote automation control, and IoT data. For each of those types of traffic, QoS requirements are identified (bandwidth, delay, priority), and DSCP marking and 5G traffic priorities are associated. A DT use-case scenario is presented, illustrating the applicability of the model. 7. P. Dhungana (Department of Electrical and Computer Engineering, University of Kentucky, Lexington, United States), D. Vinogradac, Z. Korman, M. Špoljarić, J. Balen (Faculty of Electrical Engineering, Computer Science and Information Technology, Osijek, Croatia) A Distributed Resource Sharing Mechanism for Enhanced Edge Computing in VANETs  Dynamic architecture, diverse mobility speeds of nodes, and its variations in forming sophisticated structures from a simple structure within a short period have been major challenges to overcome in the realization and optimization of the VANETs. Fog and edge computing enable VANETs to achieve a more realistic and dependable architecture of fog and cloud layers along with edges. To improve performance and optimization of the VANETs in Edge and Fog computing, collection, processing, and distribution of the information plays a vital role. Due to limited computational capabilities in the OBU, complex algorithms cannot be implemented based on data collected in the vehicle. For such applications, distributed computing could enable the application of such computationally intensive algorithms. In this paper, an algorithm for DRSM (Distributed Resource Sharing Mechanism) is developed and implemented using Veins in the OMNeT++ platform. For simulating the network and testing the algorithm, SUMO network mobility simulation is used by using randomly generated traffic conditions in the city of Erlangen, Germany. 8. A. Žaja, M. Vuković, R. Gligora, T. Kukec, M. Stuhne (Faculty of Electrical Engineering and Computing, Zagreb, Croatia) Implementation of a Distributed Mobile Network Simulator for Generating CDRs  In modern mobile networks, call detail records (CDRs) are a key data source for analyzing user behavior, monitoring network performance, and detecting unusual patterns. This paper describes the development of a distributed system that simulates the operation of a mobile network with multiple base transceiver stations (BTS). Users are identified by unique identifiers (IMEI), while a central system manages communication between base stations and generates CDRs that record user transitions between stations. The system also includes a motion simulator to generate mobile events. This solution provides a scalable environment for further research into CDR data analysis and anomaly detection. 9. F. Šklebar (Krapina University of Applied Sciences, Krapina, Croatia), V. Franki, V. Kirinčić (Faculty of Engineering, Rijeka, Croatia) A Modular ICT Architecture for Virtual Power Plants Based on Distributed Energy Resources  Virtual Power Plants (VPPs) represent a key technological concept for integrating distributed energy resources (DER) into modern power systems. Their effectiveness depends heavily on a robust, scalable and modular ICT architecture capable of supporting heterogeneous devices, diverse communication protocols and dynamic grid conditions. This paper proposes a modular ICT architecture for VPPs based on open standards, interoperable communication layers and distributed control principles. The central idea is to separate the VPP system into functionally independent ICT modules—data acquisition, device management, forecasting, optimization, and market interfacing—each capable of scaling horizontally as new DER units are added. The work analyzes the suitability of protocols such as Message Queuing Telemetry Transport (MQTT), IEC 61850, Modbus, and Open Charge Point Protocol (OCPP) for DER integration and discusses how modularity eases interoperability challenges. A conceptual system model is presented, followed by an evaluation of scalability, fault isolation, and deployment flexibility. The proposed architecture aims to provide a practical and future-proof ICT foundation for the development of reliable, distributed and market-ready VPP solutions. 13:30 - 15:00 Odmor za ručak  utorak, 26.5.2026 15:00 - 17:30, Galija, Hotel Admiral, Opatija 15:00 - 15:45 Bežični, satelitski i optički komunikacijski sustavi Predsjedatelj:   1. J. Bilandžić (Ericsson Nikola Tesla Servisi d.o.o., Osijek, Croatia) Quantum Technology in Physical Layer OSI Model  The development of optical communication networks reached physical limits, thus encouraging the research of quantum technologies at the physical layer of the OSI model. Quantum communication offers new methods for secure and efficient information transfer by exploiting the principles of quantum mechanics. Among the most mature application is Quantum Key Distribution (QKD), which enables secure communication through the transmission of cryptographic keys encoded in photon states. Quantum principles and technologies such as entanglement swapping and quantum teleportation can extend communication distances and make the foundation for future scalable quantum networks. Additional components such as beam splitters, entangled photon sources and quantum repeaters, support generation, distribution and manipulation of quantum bits (qubits) in fiber based infrastructure. Today, a key challenge lies in integrating these technologies into existing optical systems in order to make quantum networks feasible. This paper offers a review of essential principles, current implementations and integration paths of quantum technologies in physical layer, with emphasis on their potential to coexist with classical networks and establish the backbone of the future quantum internet. 2. J. Delihdodžić, J. Vuković, D. Babić (Faculty of Electrical Engineering and Computing, Zagreb, Croatia) Sky Noise Measurement for X-band LEO Satellite Communications  The distribution of sky noise across the hemisphere is essential for the development of an accurate LEO satellite communication link budget. We discuss the construction of a characterization system, and measurement results of sky noise above Zagreb, Croatia. Maps of noise density at 10.75 GHz, and 12.65 GHz were created as a function of azimuth and elevation for the entire hemisphere. These maps were used to estimate sky and antenna noise temperatures, and to show electromagnetic radiation sources in the X-band (geostationary communication satellites). 3. N. Stanojević, B. Prlinčević, B. Milosavljević, V. Simović, M. Mišić (Kosovo and Metohija Academy of Applied Studies, Leposavic, Serbia) MoE Architecture for Analyzing FSO System Performance under Atmospheric Turbulence  This paper presents a novel approach to mitigating the negative impact of atmospheric turbulence on the performance of Free Space Optics (FSO) communication systems. The proposed approach is based on the application of a Mixture of Experts (MoE) deep learning model, which enables the unification and simplification of FSO system performance analysis through a single formulation. Relevant signal quality metrics considered include the average bit error rate (ABER), outage probability (OP), and channel capacity (CC). The paper proposes the use of three statistical models to describe different turbulence intensity levels: a log-normal model for weak turbulence, a Gamma–Gamma model for moderate turbulence, and a Chi-square–inverse Gamma model for strong turbulence. These models serve as the expert components within the MoE architecture, with each expert assigned to a specific domain of atmospheric turbulence effects. In addition to the theoretical description, the paper presents a block diagram of the MoE model, which illustrates, in a simplified manner, the principle of the proposed architecture and the decision-making process for selecting the appropriate statistical model. The proposed approach enables automated identification of the dominant atmospheric turbulence effects without the need for manual model selection, thereby providing a foundation for the further development of intelligent FSO systems. 15:45 - 16:30 Bežični sustavi, elektromagnetska polja i mjerne tehnologije Predsjedatelj:   4. M. Faisal, J. Isoaho (University of Turku, Turku, Finland) Directional Dual-Antenna Design and Optimization for Physical Layer Security in Ambient Backscatter IoT  Ambient Backscatter Communication (AmBC) enables battery-free IoT devices but is highly vulnerable to eavesdropping due to its broadcast nature and limited processing capabilities. This paper proposes a dual antenna architecture to enhance physical layer security in AmBC-based IoT systems. The design combines a microstrip patch antenna array for efficient ambient signal reception and energy harvesting with a directional quasi-Yagi-Uda antenna for secure transmission. Antenna parameters are optimized using multi-objective genetic algorithm (MOGA) and multi-objective particle swarm optimization (MOPSO), applied separately to analyze their effectiveness and tradeoffs. The proposed dual-antenna system was designed and electromagnetically evaluated using CST Studio Suite, where key parameters such as S11, VSWR, gain, and radiation patterns were obtained. The antenna optimization and secrecy capacity analysis were subsequently performed in MATLAB, enabling comparative evaluation between baseline and optimized designs using MOGA and MOPSO approaches. Results show improved directional gain, reduced side-lobe levels, and enhanced secrecy capacity compared to baseline antenna designs, with clear performance differences observed between MOGA and MOPSO optimized solutions. These findings demonstrate that optimized directional antennas provide a practical and energy-efficient approach for strengthening physical layer security in AmBC-enabled IoT environments. 5. H. Maloku, E. Mucolli, K. Mustafa, J. Dobruna, M. Ibrani (University of Prishtina, Prishtina, Kosovo) Measurement and Evaluation of Magnetic Flux Density in Various Types of Vehicles  Technological advancements have led to the introduction of new innovations, including hybrid and electric vehicles. Hybrid and fully electric vehicles have become popular due to their advantages, especially based on fact that they contribute less to climate change and carbon emission. However, low frequency magnetic field exposure in electric and hybrid vehicles has raised public concern regarding their impact on human health. This exposure does not affect only the driver but also other passengers inside the vehicle. There have been many studies evaluating magnetic flux density in these vehicles, in different countries. In this paper we present a measurement-based study for evaluating the magnetic flux density levels in three different types of vehicles (hybrid, mild hybrid and electric). Measurements are conducted inside the vehicles at different locations, in urban environment for various speeds and for different operating mode. Results showed that all the measured values are well below the international guidelines. 6. M. Filipašić, M. Dadić (University of Zagreb, Faculty of Electrical Engineering and Computing, Zagreb, Croatia, Zagreb, Croatia) Comparative Analysis of Shielded Loop Magnetic Field Probes with Rectangular and Circular Shield Apertures  This paper presents a comparative analysis of two shielded one-turn magnetic field loop probes, designed in printed circuit board technology (PCB) on FR-4 substrate. The studied probes are developed with different shield aperture geometries: rectangular and circular. The influence of the shield aperture shape on probe sensitivity and electric field suppression is investigated over a wide frequency range from 1 MHz to 3 GHz. Both probe designs are analyzed using finite element method (FEM) simulations in Ansys HFSS and experimentally validated through measurements. A dedicated measurement setup is employed to evaluate the probe sensitivity and electric field suppression ratio, enabling a direct comparison between the two aperture geometries. The presented study provides useful design guidelines for optimizing shielded magnetic field probes for near-field electromagnetic interference and compatibility (EMI/EMC) measurements, with applicability as compact sensor solutions for electromagnetic radiation assessment in modern telecommunication systems. 16:30 - 16:45 Odmor  16:45 - 17:30 Modeliranje, simulacije i dizajn usmjeren na korisnika Predsjedatelj:   7. A. Periola (Cape Peninsula University of Technology , Cape Town, South Africa), A. Alonge (Tshwane University of Technology, Pretoria, South Africa), K. Ogudo (University of Johannesburg, Johannesburg, South Africa) Space Based Data Centers–System Design for Data Migration Service at End of Life   The need to realize a reduced water footprint has led to the introduction and consideration of space based data centers (SBDC). This is because SBDC cooling does not rely on earth’s water resources like the existing terrestrial data centers (TDCs). However, it is important to ensure that space based data centers (SBDCs) are able to deliver their computing services during the end-of-life approach epochs. This is important to ensure that SBDC data and algorithm execution are not lost abruptly at their end of life. The presented research addresses this challenge and proposes a distributed computing network architecture that focuses on resource constrained contexts. The proposed network architecture uses SBDC service continuity anchors (SCAs) to ensure service continuity as SBDCs approach their end-of-life. The SCAs are other SBDCs, TDCs and modular data centers (MDCs). Investigation show that the use of the proposed approach enhances the SBDC service access duration by an average of an average of (14.84– 31.23)% and (17.38–44.07)% for months 8 and 9 of SBDC degradation, respectively. 8. G. Topić (Ericsson Nikola Tesla d.d., Zagreb, Croatia) An Enhanced Colored Petri Net Framework for Dynamic Modeling of Complex Project Systems Based on DSM and DMM Models  A previously proposed theoretical approach to dynamic modeling of complex project systems, based on the conversion of Design Structure Matrix (DSM) and Multiple-Domain Matrix (MDM) models into Colored Petri Net (CPN) representations, has proven to be computationally complex, slow, and ineffective for large-scale systems with dense feedback structures. To address these limitations, this paper introduces an enhanced CPN topology together with a simplified and more transparent conversion of DSM and MDM models, enabling near-instantaneous simulation performance. The proposed method employs a stationary CPN model whose topology remains invariant with respect to project-specific input parameters. Input data, declaratively specified from DSM and MDM representations, are processed by the CPN model exclusively as a mathematical–process simulation mechanism, without requiring structural modifications for individual project instances. The model efficiently incorporates stochastically defined feedback relationships, enabling dynamic behavior analysis of complex project systems.The implementation leverages advanced features of Colored Petri Nets, including CPN Markup Language (CPN ML) for data preparation, hierarchical net structures, and extended modeling constructs available in recent versions of CPN Tools. 9. L. Vučen, M. Kraševac, A. Radović, I. Slošić (Faculty of Electrical Engineering and Computing, Zagreb, Croatia) What If You Had to Communicate Differently? Teaching AAC Through Interactive Design  Augmentative and Alternative Communication (AAC) refers to various methods used by people with communication difficulties to express themselves besides speaking, including gestures, manual signs, and a range of low-tech and high-tech assistive technologies. Although AAC plays a crucial role in daily communication for many individuals, public awareness remains limited, often creating barriers to social inclusion. This paper presents the design and development of an educational application, i.e. a serious game, intended to raise public awareness of AAC and familiarize users with the basic principles of AAC use, focusing on symbols, communication boards, and switches, which are assistive devices that support alternative navigation methods. The serious game follows Universal Design for Learning principles, ensuring accessibility, responsiveness, and ease of use for people with diverse abilities and experience levels. It is designed to implement sequential learning and allow users to apply acquired knowledge through practical tasks across four levels. Serving as a proof of concept, the developed serious game demonstrates the value of its educational objectives and highlights potential avenues for future enhancements in design and user experience. By combining learning, interactive design, and game elements, this solution aims to increase understanding, raise awareness, and support more inclusive communication in everyday life.

Osnovni podaci:

Željka Car (Croatia), Stjepan Golubić (Croatia), Dragan Jevtić (Croatia), Branko Mikac (Croatia), Martina Antonić (Croatia)

Studentski popust se ne odnosi na studente doktorskog studija.

OBAVIJEST AUTORIMA: Uvjet za objavu rada je plaćanje najmanje jedne kotizacije po radu. Autorima 2 ili više radova, ukupna se kotizacija umanjuje za 10%.

Željka Car 
Fakultet elektrotehnike i računarstva
Unska 3
10000 Zagreb, Hrvatska

Tel.: +385 1 6129 787
Mobilni tel.: +385 91 507 3452
E-mail: zeljka.car@fer.hr
 

Najbolji radovi bit će nagrađeni.
Prihvaćeni radovi bit će objavljeni u zborniku radova s ISSN brojem. Radovi na engleskom jeziku prezentirani na skupu bit će poslani za uključenje u digitalnu bazu IEEE Xplore.

Opatija je vodeće ljetovalište na istočnoj strani Jadrana i jedno od najpoznatijih na Mediteranu. Ovaj grad aristokratske arhitekture i stila već više od 180 godina privlači svjetski poznate umjetnike, političare, kraljeve, znanstvenike, sportaše, ali i poslovne ljude, bankare, menadžere i sve kojima Opatija nudi svoje brojne sadržaje. 

Opatija svojim gostima nudi brojne komforne hotele, odlične restorane, zabavne sadržaje, umjetničke festivale, vrhunske koncerte ozbiljne i zabavne glazbe, uređene plaže i brojne bazene i sve što je potrebno za ugodan boravak gostiju različitih afiniteta. 

U novije doba Opatija je jedan od najpoznatijih kongresnih gradova na Mediteranu, posebno prepoznatljiva po međunarodnim ICT skupovima MIPRO koji se u njoj održavaju od 1979. godine i koji redovito okupljaju preko tisuću sudionika iz četrdesetak zemalja. Ovi skupovi Opatiju promoviraju u nezaobilazan tehnološki, poslovni, obrazovni i znanstveni centar jugoistočne Europe i Europske unije općenito.

Detaljnije informacije se mogu potražiti na www.opatija.hr i www.visitopatija.com.