Collaborative Research: VINES: Track 2: Heterogeneous Edge Intelligence for Natural Disaster Response Analytics (HEIDRA)

While natural disasters remain a significant threat to human life and economic stability, emergency responders navigating these unfamiliar, harsh environments are often equipped with outdated technologies that fall short of accurately and rapidly locating survivors and assessing damage. However, deploying advanced artificial intelligence (AI)-based technologies, including unmanned vehicles, to augment disaster response strategies in limited-connectivity environments requires overcoming critical challenges. In particular, unreliable networks, limitations in device capabilities, the inability to effectively prioritize vast and diverse data streams, and the lack of integration of human intuition into automated systems stymy AI and network operators from supporting the nation’s most urgent public safety needs. The HEIDRA (Heterogeneous Edge Intelligence for Natural Disaster Response Analytics) partnership of academia, industry, and disaster response stakeholders will deliver a resilient and reliable decision-making solution for emergency missions, using at-scale, live missions to validate enabling mobile and multi-access edge computing technologies under real-world conditions. Adoption-ready technology will scale to broader public safety efforts such as search and rescue by providing trustworthy next generation (NextG) radio efficiency, the first-time ability to rapidly adapt and reallocate resources, and local on-site, on-device intelligence. HEIDRA outcomes will impact future networking technology standards, while industry-relevant training materials and outreach will accelerate workforce development and commercialization across the emergency response and public safety sectors. HEIDRA (Heterogeneous Edge Intelligence for Natural Disaster Response Analytics) delivers a transformative approach to mobile/multi-access edge computing (MEC)-augmented disaster response by solving four principal challenges: (i) unreliability of local access and backhaul infrastructure in disaster zones; (ii) size, weight, and power constraints of first responder assets; (iii) latency sensitivity of emergency services; and (iv) the need for accurate and intuitive on-device analytics. HEIDRA achieves a high-fidelity, human-in-the-loop, real-time, dynamic intelligence solution for NextG public safety by innovating and integrating three complementary MEC enabling technologies: (1) distributed and collaborative AI/ML (Machine Learning) training and inference across heterogeneous edge assets; (2) importance-based mobile resource management and task-oriented network planning; and (3) responder-in-the-loop decision-making and analytics feedback. The full-stack HEIDRA software platform will be deployed on edge sensing and computing hardware, supported by environment-hardened communication, networking, and AI/ML protocols along with a fine-tuned mobile graphics user interface (GUI). HEIDRA will be field-tested end-to-end through live missions conducted by first responders with mobile phones, unmanned vehicles, and other portable nodes in an at-scale emergency management training center emulating natural disaster search and rescue missions. 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.