Agentic Semantic Communication: Concepts, Architectures, and Visions

Abstract

Sixth-generation (6G) applications are increasingly requiring receivers to execute tasks based on received information. However, existing approaches optimize bit-level reconstruction independently of task relevance, so scarce transmission resources can be spent on task-irrelevant information, making the resulting data rate requirements increasingly difficult to meet. This discrepancy between the purpose of communication and the objective being optimized has led to a paradigm shift toward semantic communication (SC). Although prior surveys have examined SC, autonomous SC systems capable of adaptability and reasoning remain insufficiently explored. To address this research gap, we present a concept of Agentic SC (ASC) which emphasizes autonomy by enabling agents in communication systems to dynamically perceive, plan, reason, and self-correct in response to tasks and environmental changes. In this survey, we establish the conceptual foundation of ASC and propose a new taxonomy of SC based on the degree of system-level integration. Within this taxonomy, we review existing studies, highlight representative techniques, and draw lessons across different lines of work. Specifically, we explore how agents based on large-scale language models, world models, external knowledge, orchestration, and emergent communication can achieve autonomy and adaptability. In addition, we investigate collaborative communication and coordination strategies, as well as edge deployment techniques that enable ASC systems to operate at scale. Furthermore, we identify promising applications and use cases for ASC, identify open challenges for system implementation, and suggest future research directions.