Collaborative Research: VINES: Track 1: Toward Agentic AI-powered NextG Network with Integrated Sensing and Communication for IoT-based Healthcare

Current wireless networks face significant challenges in meeting the ultra-low-latency, high-bandwidth, and guaranteed-reliability requirements for supporting real-time healthcare. This project develops intelligent, adaptive next-generation wireless networks to meet these requirements by integrating sensing and communication with a hierarchical, agentic artificial intelligence architecture. It turns Internet-of-Things (IoT) devices and AI-enabled routers into health monitors. These monitors can detect falls, seizures, and other urgent health events and send alerts instantly. The architecture leverages AI agents working together across the network to make fast and robust decisions. The proposed solution will provide foundational support for real-time healthcare monitoring systems to enhance public health outcomes, support proactive disease management, and reduce hospitalizations and healthcare costs. The education program includes curriculum development and the involvement of students in research on wireless networking and healthcare-related IoT systems. This project aims to design a novel agentic AI-enabled wireless network architecture with multimodal integrated sensing and communication to support the growing demand for real-time healthcare. The proposed research includes developing a hierarchical agentic artificial intelligence-powered network architecture for healthcare that integrates generative AI to enable interpretable, multi-step decisions. Furthermore, a multimodal integrated sensing and communication system is designed to address joint self-interference and cross-band interference, with health-centric domain adaptation mechanisms for rapid model adaptation to real-world dynamics. Additionally, an application-aware routing layer with an agent-driven service mechanism is developed to optimize the distribution of healthcare services across edge servers. Resource-aware agents and a tiered inference strategy are further designed to deploy mixture-of-experts models efficiently on edge servers. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.