Semantic Communication for Edge Intelligence: Theoretical Foundations and Implications on Protocols

Abstract

Semantic communication has recently attracted considerable attention, mainly motivated by the trend of developing 'task-oriented' communication solutions that tailor resource consumption to the task at hand. Despite the general intuition that semantic communication may contribute to more efficient system design, there have been only a few concrete attempts to implement aspects of it in practice. To help bridge this gap, in this paper, we revisit the theoretical foundations of semantic communication and address the possible implications on the protocol and system design. The focus is on two perspectives of semantic communication: (i) a goal-oriented perspective, which unifies aspects of traffic generation, communication, and control, with emphasis on the definition of appropriate semantic-aware metrics, and (ii) a semantic operability perspective, which extends the notion of data exchange through standardized interfaces (interoperability), to include the meaning or, more generally, the significance of data. We discuss applications of the concepts in scenarios such as robotic control and health monitoring.