Authors - Tanmay M S, Neeraj Rajiv Shivam, Veena Shivanna, Sathya D, Chandramouleeswaran Sankaran Abstract - This paper presents the development and deployment of an AI-powered accident detection system for scooters, leveraging edge computing and machine learning for real time crash and fall detection with an alert system. The system architecture is divided into three layers. Layer 1 focuses on sensor data acquisition in real-time, integrating an MPU6050 sensor (accelerometer and gyroscope) with an ESP32 microcontroller for localized data processing; it’s also equipped with a GPS module to provide location data to improve emergency response. Layer 2, AI-based accident detection layer, uses a sequential neural network that is trained on diverse datasets that comprises of crash scenarios, fall scenarios and internally collected normal riding. Layer 3, the decision and notification layer, deploys real-time alerts communicating critical information such as name,blood group and GPS coordinates to emergency services and personal contacts.The proposed system combines machine learning and edge computing to provide a scalable real-time solution for improving the safety of scooter riders.
