Researchers at The University of Texas at San Antonio have created a self-sustaining flood warning system powered by artificial intelligence. The system, developed by assistant professor Chen Pan in the electrical engineering department, uses solar energy, multiple environmental sensors, long-range wireless communication, and on-device machine learning. The project also involves Mimi Xie, an assistant professor of computer science, and collaborators from Texas A&M University-Corpus Christi. Unlike traditional flood monitoring systems that rely on external power and internet connections, this new system operates independently. It uses sensors to measure temperature, humidity, light, precipitation, and water levels at different heights. By analyzing multiple factors at once, the system offers a more accurate and reliable assessment of flood risk compared to devices that rely on a single measurement. A key innovation is the use of TinyML, a specialized area of computer science that allows machine learning algorithms to run directly on small, low-power microcontrollers. This on-device artificial intelligence processes sensor data to predict imminent flooding without needing a central server. The system has shown high accuracy in early tests, achieving 98.82% validation accuracy after training. The system is powered by a solar energy harvesting system that charges a built-in battery using ambient light. This allows the device to operate indefinitely with minimal energy, far less than what a smartphone would use. When flood risk is detected, the system sends low-power radio signals using LoRa (Long Range) technology. These signals can travel over half a mile in urban areas or up to five miles in open spaces without requiring cell towers or power lines. At a central monitoring location, custom software collects data from multiple devices across a region. A more powerful AI model then analyzes this data to create interactive maps that show flood patterns over time. These maps help emergency responders make timely decisions, such as closing roads or issuing warnings before flooding reaches its peak. The cost of building one prototype device is between $150 and $220. The team is working to refine the design into a commercial product that can be used by small coastal towns, homeowners' associations, and other organizations at risk of flooding.