Smart Firefighting Early Warning and Evacuation System Based on Edge Computing with Led Dynamic Indicator Devices

The current standard practice for building fire safety relies heavily on regular lighting, emergency lighting, and smoke sensors. The management and control centers for these systems are typically separate and independent, resulting in a bulky and expensive local control system that is difficult to use and manage. Additionally, the traditional fixed instructions in fire protection systems do not allow for adaptive planning of optimal escape routes based on the location of the fire, leading to delays in evacuation and ultimately causing significant casualties. This paper presents the design of a smart fire warning and evacuation system based on edge computing utilizing LED dynamic indication devices. The system integrates LED dot-matrix technology to create a smart indicator and emergency light with dynamic adaptive indication capabilities. Additionally, it includes temperature, humidity, and smoke sensors to monitor the building's environmental conditions. Through Mesh wireless networking, time-series sensing data is aggregated to the edge gateway, where a miniature machine learning model trained by the Edge Impluse (EI) platform performs edge computing for predicting and reasoning about fire situations. The data is then uploaded to a cloud server and forwarded to mobile users via Cat.1 communication protocol for real-time monitoring. The system also incorporates an improved ant colony algorithm for automatic execution in the cloud and allows manual commands from mobile terminals to control adaptive and dynamic indication of indicator lights and emergency lights, guiding building personnel to the safest evacuation path. System testing demonstrates a high accuracy rate of 95% for fire prediction and inference, with an average inference time of less than 1 ms. This system provides a novel approach to adaptive building fire prediction and evacuation, enhancing safety measures for orderly and secure evacuations.