CNN-GA Approach for Priority Recognition and Traffic Signal Optimization in Urban Emergency Response

Abstract

Urban congestion often delays emergency services, thus endangering lives in critical situations. In this work, we propose an innovative system that combines Convolutional Neural Networks (CNNs) with Genetic Algorithms (GAs) to improve the prioritization of emergency vehicles in smart urban environments. Based on real-time traffic data from the Internet of Things (IoT), the system detects emergency vehicles (EVs) using visual cues like flashing beacons and vehicle shapes. The model was trained on varied global datasets and tested specifically in Moroccan urban environments, with an accuracy rate of 82.6% in emergency vehicle detection in varied environments. Meanwhile, the GA-based optimization also improves traffic signal control, reducing average delays at intersections by 37% in simulated tests while maintaining balanced prioritization between emergency and normal traffic. Designed for adaptability, the framework operates effectively in cities with heterogeneous infrastructures by dynamically adjusting to local traffic patterns. The results demonstrate that our approach significantly improves emergency response times and enhances traffic efficiency in complex urban settings. By applying this model in real-world contexts in Morocco, where infrastructure variability poses unique challenges, we can contribute to ongoing efforts toward harmonizing technological progress with public safety, developing a scalable model for the build-out of smarter and more resilient cities.