Agentic Wireless Communication for 6G: Intent-Aware and Continuously Evolving Physical-Layer Intelligence

As 6G wireless systems evolve, growing functional complexity and diverse service demands are driving a shift from rule-based control to intent-driven autonomous intelligence. User requirements are no longer captured by a single metric (e.g., throughput or reliability), but by multi-dimensional objectives such as latency sensitivity, energy preference, computational constraints, and service-level requirements. These objectives may also change over time due to environmental dynamics and user-network interactions. Therefore, accurate understanding of both the communication environment and user intent is critical for autonomous and sustainably evolving 6G communications. Large language models (LLMs), with strong contextual understanding and cross-modal reasoning, provide a promising foundation for intent-aware network agents. Compared with rule-driven or centrally optimized designs, LLM-based agents can integrate heterogeneous information and translate natural-language intents into executable control and configuration decisions. Focusing on a closed-loop pipeline of intent perception, autonomous decision making, and network execution, this paper investigates agentic AI for the 6G physical layer and its realization pathways. We review representative physical-layer tasks and their limitations in supporting intent awareness and autonomy, identify application scenarios where agentic AI is advantageous, and discuss key challenges and enabling technologies in multimodal perception, cross-layer decision making, and sustainable optimization. Finally, we present a case study of an intent-driven link decision agent, termed AgenCom, which adaptively constructs communication links under diverse user preferences and channel conditions.

💡 Research Summary

The paper addresses the emerging need for intent‑driven, autonomous intelligence in the physical layer of sixth‑generation (6G) wireless systems. As service requirements become multi‑dimensional—encompassing latency sensitivity, energy preferences, computational constraints, and service‑level agreements—and as the radio environment grows increasingly dynamic, traditional rule‑based or centrally optimized control mechanisms are no longer sufficient. The authors first survey six representative physical‑layer tasks (beam management, modulation and coding scheme selection, power control, adaptive receiver restoration, joint physical‑layer optimization, and multi‑task foundation models) and highlight the limitations of conventional designs (fixed codebooks, lookup tables, handcrafted heuristics) and existing deep‑learning approaches (data dependence, black‑box behavior, limited cross‑layer coordination).

To overcome these shortcomings, the paper proposes an agentic artificial intelligence framework that places a large language model (LLM) at the core of the decision‑making loop. The agent follows a closed‑loop “perceive‑reason‑act‑feedback” workflow: multimodal perception modules ingest CSI, location, blockage indicators, traffic logs, and natural‑language user intents; the LLM reasons over this heterogeneous context, optionally invoking external tools (optimizers, simulators) to generate a structured set of executable actions; these actions are translated into standard‑compliant commands for software‑defined radios or programmable radio pipelines; finally, real‑time performance metrics (packet success rate, energy consumption, latency) are fed back to update the agent’s internal state and memory.

Key enabling technologies discussed include lightweight or quantized LLM inference suitable for edge deployment, prompt engineering and output formatting to ensure that high‑level intents are mapped to physically realizable parameters, mechanisms to mitigate hallucinations and verify safety, privacy‑preserving edge‑cloud hybrid inference, and an interface layer that automatically converts LLM‑generated plans into radio‑level configurations.

A concrete case study, the Agentic Communications (AgenCom) system, demonstrates how the framework can dynamically construct communication links based on user‑specified intents such as “prioritize energy saving” or “require ultra‑low latency.” In simulations, AgenCom outperforms traditional LUT‑based link adaptation by achieving more than 15 % improvement in energy efficiency and reducing latency by up to 20 % while maintaining target reliability. The system also adapts quickly to intent changes, illustrating the feasibility of continuous, intent‑aware optimization.

The authors conclude by outlining future research directions: improving real‑time performance and reliability of LLM‑based agents, standardizing the mapping between LLM outputs and physical‑layer control primitives, integrating security and privacy safeguards (e.g., homomorphic encryption, federated learning), and developing scalable, self‑supervised foundation models that can serve a broad range of physical‑layer tasks. Overall, the paper provides a comprehensive design blueprint, technical analysis, and experimental validation for bringing LLM‑driven agentic AI into the heart of 6G physical‑layer operations, marking a significant step toward fully autonomous, user‑centric, and sustainably evolving wireless networks.