Automated Emergency Response System

Abstract

Road accidents remain the primary global cause of death since emergency responders take too long to reach accident sites which results in unnecessary fatalities. This paper presents the Automated Emergency Response System (AERS), a novel AI-powered, IoT-integrated platform that fully auto- mates the accident detection-to-dispatch pipeline. The system combines two independent detection systems which include (1) a computer vision pipeline that uses a fine-tuned MobileNetV2 deep neural network to detect accidents through real time video monitoring and (2) an ESP32 micro- controller which uses an MPU6050 inertial measurement unit (IMU) to detect physical crashes through accelerometer and gyroscope threshold measurements. The system uses Haversine distance algorithm for backend detection to find the closest ambulance while the system uses OSRM API to determine the driving path and it communicates with traffic lights along the path and it sends GPS-enabled SMS notifications to drivers through Twilio and it creates a real-time command center dashboard. The system provides deployment options for both local use and cloud services through Render.com while it connects with seven different APIs to provide geolocation services and routing functionalities and mapping capabilities and classification services. The experimental results show that accident detection achieves reliable performance through a 15-frame consensus method which eliminates false positives and the system enables dispatch within one second while using OpenStreetMap’s Overpass API to find operational resources. The proposed architecture enables emergency dispatch systems to operate with less human control while providing a scalable emergency management solution through an open source platform for smart city environments.

Introduction

The text discusses the growing severity of road traffic accidents and the urgent need for faster, automated emergency response systems. According to WHO, millions are injured and over 1.35 million people die annually in road accidents, largely due to delays in traditional emergency response processes that depend on manual reporting, coordination, and ambulance dispatch.

To solve this, the paper proposes an Automated Emergency Response System (AERS) that uses modern technologies like deep learning, IoT devices, and cloud computing. The system detects accidents using two methods: a computer vision model based on MobileNetV2 analyzing video feeds, and an IoT-based crash detection module using ESP32 with an MPU6050 sensor to identify physical collision patterns. This dual approach improves accuracy and reliability.

Once an accident is detected, the system automatically triggers emergency actions such as selecting the nearest ambulance, generating optimized routes, sending alerts, and notifying authorities via SMS. A web-based command center provides real-time ambulance tracking and coordination using tools like Flask, Leaflet.js, OSRM, and OpenStreetMap APIs.

Conclusion

The Automated Emergency Response System (AERS) shows its complete system which utilizes artificial intelli- gence and Internet of Things technology to automatically detect road accidents and send emergency responders to the scene. The system develops its first innovative feature through its ability to combine computer vision technology (which uses MobileNetV2 as its base and 15-frame consensus filter) with physical crash detection systems (which use ESP32 and MPU6050 IMU). After the system confirms an accident detec- tion, it connects to the closest ambulance and hospital through Haversine-based nearest-neighbor search. The system executes all tasks, including driving route calculation, traffic signal automation, and GPS embedded SMS notification dispatch, within several seconds. AERS uses its dual local-cloud deployment architecture to operate in different environments which range from simple demonstration systems to complex cloud-based operations that cover multiple sites. The system gains real-world geographic data and dependable communication capabilities through its integration with seven external APIs (OSRM, Over-pass, Twilio, Roboflow, Google Geolocation, TinyURL, IP-API). The experimental evaluation results show that AERS system achieves 92% validation accuracy for accident detection while reducing false positive rates by 85% through its consensus mechanism and providing dispatch times below three seconds for complete system operation. The results demonstrate how integrated AI IoT systems can decrease emergency response times, which leads to increased life-saving potential. The complete system from firmware to backend and fron- tend code has research and deployment access. This system serves as the base for developing future smart city emergency management systems.

References

[1] World Health Organization, “Global status report on road safety 2023,” WHO Technical Report, Geneva, Switzerland, 2023. [2] F.-H. Tung, J. Zelek, and D. Clausi, “Goal-based trajectory analysis for unusual behaviour detection in intelligent surveillance,” Image and Vision Computing, vol. 29, no. 4, pp. 230–240, 2011. [3] W. Sultani, C. Chen, and M. Shah, “Real-world anomaly detection in surveillance videos,” in Proc. IEEE/CVF Conf. Computer Vision and Pattern Recognition (CVPR), Salt Lake City, UT, USA, 2018, pp. 6479– 6488. [4] A. Mohan, S. Poobal, “Accident detection using convolutional neural networks,” Computers & Electrical Engineering, vol. 96, p. 107444, 2021. [5] H. Chen, S. Iyer, G. Chen, and W.-S. Lu, “CarSafe app: alerting drowsy and distracted drivers using dual cameras on smartphones,” in Proc. ACM MobiSys, Florence, Italy, 2015, pp. 13–26. [6] R. Kumar and S. Palanisamy, “IoT-based accident detection and alert system using ESP32 and MPU6050,” International Journal of Recent Technology and Engineering, vol. 8, no. 3, pp. 5212–5216, 2019. [7] A. Rajagopalan, V. Bhatia, and M. Sharma, “GIS-based optimal am- bulance routing for emergency medical services,” Journal of Ambient Intelligence and Humanized Computing, vol. 11, pp. 2911–2925, 2020. [8] D. Luxen and C. Vetter, “Real-time routing with OpenStreetMap data,” in Proc. 19th ACM SIGSPATIAL Int. Conf. Advances in Geographic Information Systems, Chicago, IL, USA, 2011, pp. 513–516. [9] T. Tamrat and N. Kachnowski, “Special delivery: an analysis of mHealth in maternal and newborn health programs and their outcomes around the world,” Maternal and Child Health Journal, vol. 16, pp. 1092–1101, 2012. [10] M. Sandler, A. Howard, M. Zhu, A. Zhmoginov, and L.-C. Chen, “MobileNetV2: inverted residuals and linear bottlenecks,” in Proc. IEEE/CVF Conf. Computer Vision and Pattern Recognition (CVPR), Salt Lake City, UT, USA, 2018, pp. 4510–4520.

Copyright

Copyright © 2026 Shriya Malode, Pranav Pawar, Dr. M. P. Sardey. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.