Generative AI Meets Wireless Networking: An Interactive Paradigm for Intent-Driven Communications

This paper introduces an innovative paradigm in interactive wireless networking, leveraging large language models (LLMs) and generative artificial intelligence (GenAI) to dynamically align network configurations and transmission strategies with user intents. Central to this paradigm is a “human-in-the-loop” framework that incorporates two pivotal processes: intent alignment with the network and streamlined user interaction. Existing quality of experience (QoE) modeling methodologies often suffer from a weak correlation between user intent and network adjustments. To address this, we review intent consistency in QoE modeling, utilizing reinforcement learning from human feedback (RLHF) and advanced prompt engineering to seamlessly integrate human intent into network configurations. Surrounding our paradigm, we provide a comprehensive survey on a multi-layered integration that spans the application, intent, semantic, network, and transmission layers, seamlessly translating user intent into optimized networking and communication outcomes. We investigate key technologies such as semantic communication and GenAI-enabled transmission and networking for pioneering applications, which allow for the real-time understanding and execution of high-layer intents, enabling the integration of user intents with wireless networks. Additionally, we explore cloud-edge-device collaboration within our paradigm, which distributes computational and semantic tasks across the network infrastructure to enable low-latency, scalable, and intelligent interactions. The paper concludes with an analysis of existing challenges and prospective research directions, highlighting the transformative potential of GenAI in shaping the future of intent-driven wireless networks.