Twinning action for spreading excellence in Artificial Intelligence of Things

We are excited to share that our joint research paper, "Reducing Communication Overhead in the IoT-Edge-Cloud Continuum: A Survey on Protocols and Data Reduction Strategies" , has been published in the Internet of Things Journal (Elsevier) and is also available on arXiv . 🔗 Read the journal article 🔗 View the arXiv preprint This work is the result of a collaboration between Dora Kreković, Petar Krivić, Ivana Podnar Žarko, and Mario Kušek from the University of Zagreb Faculty of Electrical Engineering and Computing (FER) , and Danh Le Phuoc from Technische Universität Berlin (TU Berlin) , reflecting a strong collaborative effort between the two institutions. The paper offers a comprehensive survey of strategies to reduce communication overhead in the IoT-edge-cloud continuum. It provides: A detailed overview of state-of-the-art techniques for minimizing communication overhead A qualitative evaluation and comparison of communication protocols and data reduction strategies A roadmap for developing scalable, efficient IoT solutions capable of real-time processing while mitigating network congestion Insights into integrating protocols and reduction strategies across the layers of the IoT-edge-cloud continuum, with attention to the constraints posed by resource-constrained IoT and edge devices This research contributes to the advancement of efficient, communication-aware IoT system design and offers practical guidance for both academic and industry stakeholders.

We're excited to share that Prof. Ivana Podnar Žarko, AIoTwin coordinator, presented the AIoTwin Orchestration Middleware at the STEADINESS workshop during HiPEAC 2025 in Barcelona (Jan 20-22, 2025)! 🚀 Her presentation, " Artificial Intelligence of Things (AIoT): Bringing Intelligence into the Physical World ," highlighted the critical need for adaptive and QoS-aware orchestration middleware in the evolving IoT-edge-cloud continuum. This middleware, now open-source on our GitHub repository https://github.com/aiotwin , is designed to manage both learning and inference pipelines. Prof. Podnar Žarko presented the requirements and architecture of our solution, which extends traditional orchestrators like Kubernetes with AIoT-specific capabilities. We're focused on optimizing resource utilization by considering both infrastructure-related QoS (latency, communication cost) and AI-related performance metrics (accuracy, loss). Key takeaways from the STEADINESS workshop emphasized the necessity for: scalable and adaptable engineering methodologies, especially in safety-critical sectors like automotive, aerospace, and industrial automation; modular and interoperable solutions leveraging RISC-V and AI for embedded systems to enhance resilience and performance. clear AI regulation guidelines that balance innovation with ethical and security considerations. For more information, check out the workshop's joint whitepaper: https://zenodo.org/records/14920027 .

Mislav Has participated in the fourth short-term staff exchange as part of the AIoTwin project, hosted by the Research Institutes of Sweden, RISE  from November 19 to December 18, 2024.  During this period, Mislav collaborated closely with researchers at RISE, exploring advanced techniques for deploying WebAssembly on constrained devices and addressing key challenges in balancing performance and resource efficiency. The exchange involved in-depth discussions, hands-on experimentation, and the initiation of joint research efforts in the field of IoT and energy-efficient computing. The outcomes of this collaboration will be disseminated through future publications, paving the way for ongoing partnerships and advancements in IoT technologies.  For those interested in a visual recap of this exchange, a YouTube video documenting highlights is available  here .  

Dr. Fehmi Ben Abdesslem visited the University of Zagreb Faculty of Electrical Engineering and Computing between November 12th and 14th, 2024, as part of the AIoTwin project. His visit contributed to strengthening collaboration and knowledge sharing within the project.  On November 12th, Dr. Ben Abdesslem delivered a public talk titled "Battery-Free IoT" , organized by the Croatian Section of IEEE. The talk, attended by members of the IoTLab involved in the AIoTwin project, sparked engaging discussions on the potential and challenges of energy-efficient, battery-free IoT solutions. In the following two days, Dr. Ben Abdesslem led two interactive sessions for the IoTLab members. The first session, Lecture on WebAssembly for IoT , introduced the fundamentals of WebAssembly and its potential applications in the IoT domain. The second session, Hands-On Practice of WebAssembly for IoT , provided participants with practical experience in creating, compiling, and deploying WebAssembly modules for IoT devices. Both sessions were streamed online for the wider AIoTwin consortium, enabling broader participation. More information about the visit, can be found here .

Ivan Kralj participated in the third short-term staff exchange within the AIoTwin project. The exchange took place from April 10 to May 9 2024 at Research Institutes of Sweden, RISE . During this period, Ivan collaborated with RISE researchers to tackle cutting-edge challenges in decentralized machine learning and Spatio-Temporal Graph Neural Networks (ST-GNNs). The results of this exchange will be presented through a forthcoming conference and journal articles, building a strong partnership with RISE and advancing research in distributed ST-GNN training. For those interested in a visual recap of this exchange, a YouTube video documenting highlights and key findings is available below:

At the 49th IEEE Conference on Local Computer Networks (LCN) , to be held from October 8th to 10th, 2024, in Caen, Normandy, France, Anna Lackinger will present the paper titled "Inference Load-Aware Orchestration for Hierarchical Federated Learning" on October 9th. The paper is co-authored by Anna Lackinger, Pantelis A. Frangoudis, Ivan Čilić, Alireza Furutanpey, Ilir Murturi, Ivana Podnar Žarko, and Schahram Dustdar. This research is the result of a collaboration between the Distributed Systems Group (DSG) at TU Wien's Institute of Information Systems Engineering and the IoTLab at FER. The paper addresses the challenge of orchestrating machine learning workflows composed of hierarchical federated learning (FL) based training and model serving processes.

body { font-family: Arial,sans-serif; background-color: #fff; } #cms_area_horizontal_nav .navbar-nav>.nav-element>a { padding-top: 28px; color: rgba(15, 108, 177) !important; } #cms_area_horizontal_nav { background-color: rgba(255,255,255) !important; } .login_icon a { color: #fff !important; } .login_icon { background-color: rgba(15, 108, 177); } nav a { color: rgba(15, 108, 177) !important; } .headermenu > #cms_area_horizontal_nav { background-color: rgba(255,255,255) !important; } } The Horizon Europe twinning project AIoTwin organised its 2nd Summer School on the intersection of Internet of Things (IoT) and Artificial Intelligence (AI) in September 2024 in the enchanting city of  Dubrovnik, Croatia . The event took place at the Centre for Advanced Academic Studies (CAAS) of the University of Zagreb , which offers a stunning view of Dubrovnik's Old Town's famous City Walls and Lovrijenac Fortress. The event was organised in collaboration with the EU-funded   SmartEdge   project and brought together more than 40 participants from all over Europe. The programme of the summer school included invited talks, technical tutorials and hands-on sessions, as well as the technical programme of the workshop “ Edge AI meets Swarm Intelligence ” with paper presentations and poster sessions showcasing the latest research results by PhD students. Key highlights of the Summer School include:   Keynotes   by Aaron Ding (TU Delft) on sustainable and trustworthy edge AI for future computing, Stefan Nastic (TU Vienna) on computing paradigms for next-generation computing landscapes, Spyros Lalis (University of Thessaly )  on elevating drones to first-class citizens in the cloud-edge-IoT continuum.  Tutorials on doctoral education at KTH, insights on operating distributed computing continuum systems through active inference, exploring privacy-preserving computation techniques for edge AI, in-network programming and practical guidelines for startups with examples. Hands-on sessions included two open-source solutions: 1) AIoTwin orchestration middleware enabling edge orchestration of hierarchical Federated Learning workflows and 2) ColonyOS , an event-driven meta-operating system spanning edge and cloud environments. Joint technical workshop with the project SmartEdge on “ Edge AI meets Swarm Intelligence ”. The event successfully promoted collaboration and networking between academics, PhD students and industry professionals through interactive sessions, giving researchers and participants the opportunity to share insights on edge AI , swarm intelligence and edge orchestration . More information about the event with talk abstracts and slides can be found here , and photos from the event are included below ("More").  

The project leader and head of the Internet of Things Laboratory at FER, Prof. Ivana Podnar Žarko , delivered a talk titled "Artificial Intelligence of Things (AIoT): Bringing Intelligence into the Physical World" at the Digital Futures cross-disciplinary research center in Stockholm, Sweden, on August 29th 2024. The talk introduced the concept of AIoT as a continuous processing cycle consisting of four steps: sensing, learning, decision-making, and acting. This cycle was illustrated with concrete real-world use cases. The challenges of AIoT were also reviewed, highlighting its operation in a dynamic and heterogeneous environment with limited resources dedicated to real-time data processing, online learning, inference, and decision-making under time constraints. Additionally, it was noted that data streams from IoT devices are often incomplete, prone to errors, and unlabelled. The talk further emphasized that AIoT systems must adhere to strict privacy and security requirements to protect sensitive user data, ensure device integrity, and maintain the safety of the physical environment. Find out more about the event and download the talk slides here .

At the IEEE International Conference on Edge Computing and Communications - IEEE EDGE 2024 held in Shenzhen, China from 7 to 13 July 2024, the paper " QEdgeProxy: QoS-Aware Load Balancing for IoT Services in the Computing Continuum " by Ivan Čilić , Valentin Jukanović, Ivana Podnar Žarko , Pantelis Frangoudis , and  Schahram Dustdar was presented. The paper was accepted after an international review and presented remotely by Ivan Čilić. The work results from a collaboration between IoTLab at FER and DSG at TU Wien on orchestration and routing in Computing Continuum (CC). The paper introduces  QEdgeProxy , an adaptive and QoS-aware load-balancing framework specifically designed for routing client requests to appropriate IoT service instances in the CC. QEdgeProxy integrates naturally within Kubernetes, adapts to changes in dynamic environments, and manages to seamlessly deliver data to IoT service instances while consistently meeting QoS requirements and effectively distributing load across them. 

Following the success of the inaugural AIoTwin Summer School in Šibenik last year, we are delighted to announce the forthcoming second edition scheduled for September 16th to 20th 2024 , set to take place in the enchanting city of Dubrovnik, Croatia . The event will be organized in collaboration with the EU funded SmartEdge project . The summer school program includes invited talks, technical tutorials, hands-on sessions with advanced hardware, poster sessions showcasing the latest research by PhD students, and other insightful presentations and discussions. In addition, the program includes a technical workshop on “ Edge AI meets Swarm Intelligence “, featuring insightful presentations and discussions on cutting-edge research. The call for papers for the technical workshop on “ Edge AI meets Swarm Intelligence “ is open by June 15th, 2024. Join us to be part of a transformative learning journey shaping the future of AIoT and Edge AI. 🔶 Logistics Dates: 16th to 20th September 2024 Venue: University of Zagreb / Sveučilište u Zagrebu Centre for Advanced Academic Studies (CAAS) 🔶 Tentative agenda 🔹 16-17 September – AIoTwin Summer School: Keynote talks, Tutorials, Hands-on sessions 🔹 18 September – Joint Sessions SmartEdge & AIoTwin: Technical Workshop on “Edge AI meets Swarm Intelligence”, Poster session, SmartEdge open-door demonstration 🔹 19-20 September – SmartEdge Summer School: Keynote talks, Hands-on sessions Further details regarding the Summer School schedule, activities and sign-up procedures will be circulated soon.