Scope

The wireless networks applications evolve from traditional human-driven and consumed applications with objective of maximizing data rates to AI-based applications for interaction of intelligent autonomous agents including machines, vehicles, and robots. Unlike humans, the autonomous agents do not require a perfect delivery of all raw bits at high data rate but rather need a proper understanding of delivered messages in the context of performed task or mission to be accomplished. This motivates to reconsider the concept of communication from bit-oriented to meaning-oriented communication networks in a resource efficient manner. Thus, the workshop targets the emerging topic of semantic communications, which facilitates reduction of communicated data and information required for a correct completion of the task or mission of autonomous agents. The semantic communications enable radio, computing, and energy efficient communication focused on delivery of only meaningful information between communicating agents. The semantic communications is also beneficial for generic agents even with multiple sensors and sources of information allowing to take a multimodality of real-life environments into account. In such case, an extraction and combination of the task/mission-related information from different sources enable further reductions in the communication burden, aiming to operate only with data beneficial for different applications. This goes hand-in hand with latest developments in generative AI enabling to recover potentially missing information on the receiver side. The workshop targets all aspects of semantic communication ranging from theoretical foundation, via resource management, system and network aspects, to integration to AI-native future networks. 

Topics of interest

The workshop will provide insight into emerging topics of semantic communication considering theory, practice, as well as standardization. Topics of interest include, but are not limited to: 

 - Theoretical foundation for semantic communication

 - Semantic representation and misalignment 

 - Semantic coding and signal processing

 - Joint source and channel coding for semantic communications

 - Semantic knowledge base construction and updating

 - Protocols for semantic communication

 - Communication and computing resource management for semantic communication

 - Integration of semantic communication and edge/cloud computing

 - Semantic communication for multi-modal data

 - Semantic communication for multi-agent or multi-user systems

 - AI/ML-driven optimization for semantic communications

 - Gen-AI for semantic communication

 - Network architectures for semantic communications

 - End-to-end semantic communication systems

 - Semantic communication for (beyond) 6G applications (autonomous systems, robotics, digital twins, metaverse, etc.)

 - Privacy and security in semantic communications

 - Experimental platforms, testbeds, and proof-of-concept for semantic communications

 - Perspectives of standardization of semantic communications

Program

14:00 - 14:10 Welcome

14:10 - 14:50 Keynote - prof. Paolo di Lorenzo: Semantic Bridges: Making AI-Native Communication Explainable and Interoperable

14:50 - 15:10 Deep Reinforcement Learning for Intelligent UAV-assisted Secured Semantic Communications

15:10 - 15:30 Computation offloading framework based on reconstruction-free semantic communication

15:30 - 16:00 Coffee break

16:00 - 16:20 Large Language Model-Based Semantic Communication System for Image Transmission

16:20 - 16:40 Toward a Unified Semantic Loss Model for Deep JSCC-based Transmission of EO Imagery

16:40 - 17:00 Secure-SFBL: Authenticating Generative Semantic Streams for Autonomous Agents

17:00 - 17:20 Spatial-Frequential Adaptive Learned Image Transmission for Precoded MIMO Systems

Abstract:  As we move toward 6G, AI-native communication promises to transform networks from bit-pipes into intelligent ecosystems that sense, reason, and exchange meaning. Yet this transformation cannot succeed on efficiency gains alone; AI systems must be able to learn semantics that are explainable, so humans can interpret and trust their behaviour, and interoperable, so heterogeneous agents can communicate and collaborate without semantic mismatch. This keynote introduces Semantic Bridges, a vision for connecting heterogeneous AI agents through aligned and interpretable semantics. We will highlight two key enablers of semantic explainability: topological methods, which uncover structural patterns in semantic representations, and token selection strategies, which clarify how semantic units are chosen and prioritized. Then, we will address the critical challenge of semantic mismatch, how AI agents with different internal semantic representations can achieve interoperability, communicate effectively, reach mutual understanding, and collaborate on complex tasks. The crux of the approach lies in semantic alignment between the latent spaces of heterogeneous AI agents. Finally, we show how this alignment extends to structural causal models, where structural semantics provide a foundation for transferring causal knowledge across models at multiple levels of abstraction. By aligning meaning between agents, semantic bridges make AI-native communication explainable, interoperable, and trustworthy.

Accepted papers

"Large Language Model-Based Semantic Communication System for Image Transmission" by soheyb ribouh, university of Rouen Normandy; Osama Saleem, INSA Rouen

"Deep Reinforcement Learning for Intelligent UAV-assisted Secured Semantic Communications" by Gaurav Kumar Pandey, Matus Dopiriak, Technical University of Kosice, Slovakia; Devendra Singh Gurjar, National Institute of Technology Silchar; Dr. Suneel Yadav, Indian Institute of Information Technology Allahabad; Juraj Gazda, Technical University of Kosice, Slovakia

"Spatial-Frequential Adaptive Learned Image Transmission for Precoded MIMO Systems" by Yiding LI, Zijian Liang, Kai Niu, Ping Zhang, Beijing University of Posts and Telecommunications

"Computation offloading framework based on reconstruction-free semantic communication" by Adam Janes, Zdenek Becvar, Jan Danek, Czech Technical University in Prague

"Toward a Unified Semantic Loss Model for Deep JSCC-based Transmission of EO Imagery" by Ti Ti Nguyen, Thanh-Dung Le, Vu Nguyen Ha, SnT, University of Luxembourg; Duc-Dung Tran, Hung Nguyen-Kha, DINH-HIEU TRAN, Carlos Luis Marcos Rojas, University of Luxembourg; Juan Merlano Duncan, Symeon Chatzinotas, SnT, University of Luxembourg

"Secure-SFBL: Authenticating Generative Semantic Streams for Autonomous Agents" by Quazi Mamun, Charles Sturt University