Semantic Communication and Computing for Scalable Edge AI

Abstract: As edge devices generate increasingly large volumes of data, executing AI tasks closer to the network edge is becoming essential to meet stringent requirements on latency, privacy, and scalability. However, this paradigm introduces challenges in communication efficiency, resource allocation, and distributed processing. This keynote explores how joint semantic communication-computation design can fundamentally improve edge AI systems. Rather than transmitting and computing raw data, semantic approaches allow to convey and process only the information that is relevant for a given task, thus substantially reducing communication and computing while preserving decision-making accuracy. We will discuss how these principles enable efficient network and split computing architectures where processing is distributed across devices and edge infrastructure. Key aspects include adaptive compression, resource-aware AI orchestration, and real-time multimodal data fusion for context-aware and scalable edge AI. Finally, we will outline open research questions in the field of semantic data communications and computing toward scalable edge AI.

Carla Fabiana Chiasserini, Ph.D.
Full Professor, Politecnico di Torino
IEEE/ACM/AAIA/WASP Fellow
Deputy Head of the Electronics and Telecommunications Dept.
Affiliated Professor with Chalmers University of Technology, Sweden