Toward Intelligent Bandwidth Prioritization in Distributed 6G ISAC: An Agentic AI Perspective

Integrated Sensing and Communication (ISAC) is a cornerstone of sixth-generation (6G) networks that is designed to offer situational awareness in real-time and intelligent coordination across increasingly sophisticated IoT environments. Yet, most existing ISAC systems are still not effective in dynamic environments, especially where semantic relevance, bandwidth-limited environments, and asynchronous data streams require adaptive, intelligent decision-making. This work provides a distributed ISAC framework integrating Agentic Artificial Intelligence (Agentic AI) with a Semantic-aware Asynchronous Federated Learning (S3A-AFL) architecture. The framework empowers edge agents to make timely context-aware decisions to dynamically allocate bandwidth, adjust sensing plans, and trigger model updates only when necessary, rather than employing pre-fixed schedules. By semantic encoding and distributed learning, S3A-AFL evades regular synchronization needs, enabling efficient performance in realistic scenarios. Compared to a static baseline, approach achieves a 32% improvement in average latency and an 18% reduction in packet loss, proving the benefit of adaptive, agent controlled. To demonstrate practical applicability, we evaluate and test the system with a storm forecasting scenario and apply a real-world environmental dataset, thereby showing how S3A-AFL enables key operations under limited network conditions. This work demonstrates the promise the framework for smart and resilient 6G applications.