SMART AUTOMOTIVE SAFETY SYSTEM USING IOT, SENSOR FUSION FOR INTELLIGENT DRIVER ASSISTANCE SYSTEM

Abstract

Road accidents remain one of the leading causes of death and serious injury, and a large share of them are linked to driver fatigue, distraction, over-speeding, unsafe following distance, and delayed reaction to obstacles. Many of the vehicles in daily use, especially budget cars, commercial vans, and older models, do not carry the advanced driver assistance features found in premium vehicles. This paper presents a Smart Automotive Safety System that uses Internet of Things technology and sensor fusion to provide intelligent driver assistance at a cost that makes retrofitting existing vehicles practical. The proposed system is built around an ESP32 microcontroller that collects data from several sensors mounted on the vehicle. Ultrasonic sensors monitor the front and rear clearance, an MPU6050 accelerometer and gyroscope module records sudden braking, sharp turns, and impacts, an eye-blink infrared sensor and a small camera observe the driver for signs of drowsiness, an MQ-3 sensor detects alcohol in the cabin air, and a GPS module provides location and speed. Readings from these sensors are combined through a sensor fusion routine so that a warning is issued only when the evidence from more than one source supports it. When a hazardous condition is detected, the system alerts the driver locally through a buzzer, a vibration motor fitted to the seat, and a small display on the dashboard. If an obstacle is approaching too quickly and the driver does not respond, the controller can reduce the throttle signal in the prototype through a relay-controlled motor driver. In the event of a collision, the impact pattern from the inertial sensor triggers an emergency routine that sends the vehicle location to registered contacts through a GSM module and pushes the event to a cloud dashboard over Wi-Fi or mobile data. All trip data, including speed profile, harsh braking events, drowsiness alerts, and alcohol readings, is uploaded to a cloud platform using the MQTT protocol. A web dashboard and a mobile application allow fleet owners or family members to view the live position of the vehicle, review past trips, and receive notifications when unsafe behaviour is recorded. This record can help fleet operators identify drivers who need additional training and can support insurance and accident investigation work. The system is intended to assist the driver rather than take control of the vehicle, and every warning is designed to be clear, short, and easy to understand while driving. Testing on a scaled prototype vehicle showed that the fusion of multiple sensors reduced false alarms compared with single-sensor triggering while keeping the response time within a fraction of a second. Future work includes integration with the vehicle CAN bus, lane departure detection using a camera, vehicle-to-vehicle communication, and the use of edge machine learning models for more reliable drowsiness detection in low light.