The AIoTwin project is pleased to announce the successful completion of the tutorial, "Orchestrating Hierarchical Federated Learning Pipelines with the AIoTwin Middleware," held in Vienna, Austria on November 18, 2025. The tutorial was part of the program at the 15th International Conference on the Internet of Things (IoT 2025).
The session, presented by project members Ivan Čilić, Ana Petra Jukić, Katarina Vuknić, and Ivana Podnar Žarko, provided participants with hands-on experience in tackling the complexities of deploying Hierarchical Federated Learning (HFL) pipelines across the Cloud-Edge-IoT continuum.
For more information, check the detailed news content.
The tutorial demonstrated how to move beyond static, centralized Federated Learning (FL) by implementing and orchestrating adaptive HFL pipelines which can be changed during a training task – new clients can join or leave a task which causes pipeline reconfiguration. Participants engaged in practical tasks, learning how HFL significantly improves communication efficiency and resilience by introducing an intermediate layer of edge aggregators.
The core focus was the practical application of the AIoTwin Orchestration Middleware for (H)FL. This open-source solution allows for adaptive, runtime management of FL pipelines, ensuring optimal performance and Quality of Service (QoS) even in dynamic and unstable edge environments.
The concepts and methodology presented in the tutorial are detailed in the accompanying open access paper:
- Čilić, A. P. Jukić, K. Vuknić, I. Podnar Žarko. Orchestrating Hierarchical Federated Learning Pipelines with the AIoTwin Middleware, Proceedings of the International Conference on the Internet of Things (IoT) Workshops 2025 1(1), 61-63. Vienna, Austria, November 18-21, 2025. doi (open access): https://doi.org/10.34749/3061-1008.2025.10
The AIoTwin consortium thanks all participants for their engagement and looks forward to further contributing to the advancement of open-source, adaptive AIoT systems.
