Abstract: We address the persistently low evacuation uptake during hydro-meteorological hazards by implementing our previously proposed “Evacuate Now Button” as a minimal, functioning system (ESP-WROOM-02 + physical button + cloud pipeline + map visualization) and conducting an initial, metrics-based evaluation. A button press wakes the device, which connects to Wi-Fi and transmits a payload; events are received and normalized on AWS Lambda, stored on Azure (Django/MySQL), and rendered with Leaflet in a two-tier arrangement. In a home Wi-Fi setting, end-to-end latency showed a median of 1.9 s for single-device trials and 2.3 s under eight nearly simultaneous presses, indicating practically responsive behavior. Across 25 simultaneous-press trials (200 presses in total), no dropped messages were observed; by the rule of three, the 95% upper bound on the loss rate is 1.5%. Power measurements of 70 mA during transmission and 10 μA in deep sleep, combined with a 2000 mAh primary cell and η = 0.85, yield estimated lifetimes of 33.0 months (once per day), 33.6 months (once per week), and 30.1 months (three times per day under higher load), comfortably satisfying seasonal operational requirements. Our contributions are: (i) a concrete bridge from simulation to functioning implementation, (ii) a reusable evaluation framework spanning latency, delivery under simultaneous presses, and battery life, and (iii) articulation of key implementation issues, including authentication, idempotency, power management, and thresholding for visualization. Limitations include the small, researcher-controlled testbed, variability in tail latency due to Wi-Fi dependence, and the need to refine privacy and operational governance. Future work will pursue quasi-experiments in field drills (with and without visualization) and reduce tail latency via SSE/WebSocket, warm functions, and communication redundancy, thereby improving external validity for real-world deployment.

Keywords: Evacuation, Agent-Based Simulation, AWS Lambda, Azure Django/MySQL, Prototype

Pages: 259-267