Collaborative Research: CPS-CIR: Community-maintained Resilient AI Infrastructure for Local Food Security in Hawaii

Modern artificial intelligence depends on large, centralized data centers, steady electricity, and constant internet access. Many rural and island communities, including farms throughout the Hawaiian Islands, cannot count on any of these. They face high energy costs, frequent weather disruptions, unreliable connectivity, and long, fragile supply chains for equipment. As a result, advanced computing tools remain out of reach in many of the places where they could do the most good, such as helping farmers make timely decisions about soil, water, and crops. This project designs small computing systems that a community can build, power, and maintain on its own. Assembled from low-cost and reused hardware such as older laptops and phones, and run on local solar power with small batteries, these locally grown systems bring useful artificial intelligence to farms without relying on distant data centers or steady internet. Farmers can ask everyday questions in plain language and receive answers drawn from their own sensors and local information, even when power or connectivity is limited. The project also gives students and agricultural workers direct experience with building and using modern computing systems. By providing farming communities dependable computing that continues to work through power and network disruptions, the project strengthens local food production and the resilience of essential infrastructure in rural and remote regions of the country, advancing the national interest in a secure food supply and reliable access to technology where centralized systems cannot reach. The project's technical goal is to design, build, and evaluate locally grown compute clusters, which are cyber-physical systems that provide modern artificial intelligence under strict limits on energy, hardware, and network connectivity. The first component is a computing cluster assembled from low-cost, mismatched, and repurposed devices such as older laptops and small computers, powered mainly by solar energy with small battery reserves. A lightweight control program monitors available power and environmental stress such as heat and humidity, allowing computation to adapt rather than assume constant electricity. The second component is a management system that distributes artificial intelligence tasks across a range of devices from small field units to larger farm and regional machines, sending each question to the smallest device capable of answering it and continuing to function when power or connectivity is lost. The system keeps its choices visible to operators, showing the tradeoffs among energy use, cost, and speed so that people remain in control of the technology. The third component is a community co-design process with beginning-farmer training programs and working farms across the Hawaiian Islands, testing whether the systems answer real agricultural questions such as soil and water management and fit existing farming practice. Together these components form a complete approach to delivering artificial-intelligence decision support in settings where reliable power and internet cannot be assumed. 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.