1.      Introduction

The Road traffic accidents a leading cause of death and injuries worldwide. The (WHO) World Health Organisation reports, each year about 1.5–1.25million people dies and 20–50 million are injured in road accidents. A critical factor in accident survival is the latency between a crash and the arrival of help. Studies show fatality risk rises with each minute of delay. Thus, automated accident detection and alert systems have been researched to overcome delays in manual response. Traditional emergency response relies on bystander calls, which can be slow [1-2]. Modern approaches leverage IoT(Internet-of-Things) devices and artificial intelligence to monitor vehicles and roadways in real time. For example, sensors (e.g. accelerometers, GPS and GSM/ GPRM) can detect collisions and immediately notify family members or police helpline services [3,4].

Traditional emergency alert systems mainly relies on manual communication or bystander help, mostly through phone calls made by bystanders or other road users. However, this approach is slow, inconsistent and not reliable, especially in rural areas highways, night time accidents or situations where the person is unconscious and unable to call for help [6]. Therefor the accident monitoring system utilise IoT to continuously monitor vehicles detect collisions and initiate emergency response automatically. By using sensors such as accelerometers, gyroscopes, GPS modules and GSM/GPRS communication modules and automatically send precise accident coordinates alert notifications to family members, police helplines, or emergency response without human help [8].

2.      Literature Review

Previous studies on accident detection cover sensor- driven vision-driven and IoT-supported techniques. Crash detection has traditionally involved accelerometers and impact sensors, within vehicles. For example Pathik et al. employ a device equipped with accelerometers, pressure sensors and GPS. Once a certain limit is surpassed the data is transmitted to the cloud where a deep learning model confirms the accident. Their framework triggered emergency notifications to hospitals maintaining false alarm occurrences by utilising ensemble CNN models. Bhatti et al. Concentrate on a smartphone centered method utilising sensors (accelerometer, gyroscope, pressure sensor, microphone) to identify collisions and environmental signals; once detected a navigation app guides responders to the location. These IoT and mobile device solutions are affordable; can be implemented in older vehicles [8-12].

Vision-based approaches examine footage from traffic cameras. For instance Shreya et al. Use a YOLO object detection model to identify vehicles and an LSTM network to detect abnormal movement patterns of an accident. They state the detector accurately identified 60% of actual accidents with high precision. Progress in real-time computer vision (such as YOLOv8) enhances both detection speed and accuracy allowing for video surveillance either onboard vehicles or, at roadside locations. These systems frequently operate on servers or edge devices constantly monitoring intersections, for crashes and issuing warnings when an abrupt deceleration or collision is detected.

The application of IoT platforms for accident reporting has been thoroughly examined. Sahraei and Al Mamari analyse research papers on driven accident detection emphasising that integrating various sensors (GSM/GPS, accelerometers, vibration) enhances accuracy. They point out that although numerous prototypes are available issues persist regarding expense, durability and communication dependability. Alkhaiwani and Alsamani suggest an IoT system that encrypts drivers’ information, during accident report transmissions underlining privacy concerns. Overall studies highlight the integration of sensors and low-delay communication to facilitate rapid emergency reaction [11].

In summary, existing literature supports real-time monitoring using IoT and AI. Sensor fusion for robust detection, and automated messaging to emergency. Our work builds on these by integrating onboard and roadside detection modalities into one unified system, with an emphasis on practical implementation and testing.

System Architecture / Design

Fig.1: System architecture of the IOT-based accident

Detection.

The suggested accident detection system employs a layered IoT framework Fig 1. Within the layer sensors are installed on each vehicle an accelerometer (to measure collision impact) and a gyroscope placed on the vehicle frame. A GPS/GSM unit captures location information. Supports mobile network connectivity. These sensors transmit data to a microcontroller ESP32 which runs algorithms to identify irregularities.

In this setup sensor signals are handled internally. For instance a sudden surge, in acceleration combined with a vibration pattern activates the crash-detection unit. At the time undergo analysis through a code to verify if a crash took place. After the system verifies an accident it gathers details GPS location .The data packet is subsequently sent on to the network layer.

The network layer is responsible for communication, where the microcontroller utilises GSM or Wi-Fi modules to transmit accident-related data to an emergency phone number in the form of an alert message. The application layer consists of a backend server that stores accident notifications and manages the emergency alert mechanism. Once accident data is received, the server logs the incident and immediately notifies the nearest emergency responders (e.g.police helpline), along with the user predefined contacts, via SMS along with location data attached.

Conceptually, the proposed system aligns with common IoT-based accident detection frameworks, where sensor fusion mechanisms detect crashes and trigger alerts to a central response system [14]. Our design enhances flexibility by supporting multiple sensor inputs. The architecture is built to ensure continuous monitoring, low false alarm rates using confirmation logic, and instant dissemination of emergency notifications, thereby improving the reliability and effectiveness of post- accident response[15-16].

3. Methodology

The conceptual Real, Time Accident Monitoring System is designed to operate in multiple layers such as the Sensor Layer, Microcontroller Layer, Network Layer, and Output Layer. The system operation is shown in Fig-1, and the entire method is detailed below.

Sensor Layer-

System is set up to keep on recording data through the following embedded sensors:

1.      Accelerometer, detects sudden deceleration/impact forces.

2.      GPS Module, acquires real, time geographic coordinates.

3.      Gyroscope, measures changes in vehicle orientation and angular motion.

By way of these sensors, the system gets direct (raw) signals about motion and location, and these signals are sent to the microcontroller for deciding (processing).

Microcontroller Layer- A microcontroller ESP32 is basically the core power unit. The signals that are received go through the three primary stages:

1. Crash Detection Algorithms:-

Working from threshold, based decision logic, it is observed that sudden changes of acceleration or orientation beyond the predetermined safety limits are checked.

2. Signal Processing:-

It does so by removing noise and ensuring that the readings are correct by using moving averages so that bumps or potholes do not cause false positives.

3. Data Verification:-

The last step in a crash is signal verification from the matching of accelerometer and gyroscope signals with the GPS timestamp data 

Network Layer-

If everything is in order, the microcontroller will send the emergency message packet through the GSM/Wi, Fi communication module.

Basically the alert packet contains:-

1. Latitude & longitude coordinates

2. Time of accident

3. Optional metadata like vehicle ID (if set up)

The GSM/WiFi layer is the one that can guarantee the delivery of the data in real, time even if there is no broadband connection.

Output Layer-

The crisis communication is divided at two points:

•            Emergency Contacts, get a detailed accident report (SMS / message with location link).

•            Police Helpline (SMS only), gets a brief emergency notification for quick localising of the problem and rapid intervention.

By using this staged transmission system, the most important facts are shared not only with relatives and friends but also with the police department.

Implementation:

The designed accident detection system is constructed with accessible and affordable hardware components. Every vehicle unit revolves around an ESP32 microcontroller acting as the controller and data processor. The ESP32 connects to an MPU6050 accelerometer/gyroscope sensor to continuously capture vehicle acceleration and positioning. A SIM800L GSM/ GPRS module handles sending emergency alerts whereas the GPS module provides real-time location data.

The MPU6050 sensor is monitored continuously. Raw acceleration data are processed and examined through threshold-driven logic. Upon detecting a slowdown or unusual force surpassing a set threshold (suggesting a possible crash) the ESP32 initiates the accident verification process. The system looks for peaks within abrief timeframe to prevent false alarms, from small bumps or potholes. The event is only classified as a confirmed accident when the threshold is reached.

Once verified the ESP32 assembles a payload that includes-

1. Vehicle ID (assigned during setup)

2. Date and time of the accident.

3.      Location coordinates (latitude & longitude from the GPS module)

Fig- 2. Illustrates the flow of the system

The initial step of implementation is the integration of various hardware elements into a small accident detection device that is installed inside the car Fig 2. The ESP32 is the main system and is connected to car’s power supply so that it can stay on all the time. The MPU6050 is fixed in a certain position inside the module so that its X, Y, and Z can detect linear acceleration and rotational movement.

The code accepted by the ESP32 is doing a continuous computation of acceleration and tilt values and always checks the current readings against the predefined accident thresholds. The limits are so set out that everyday driving, which includes the passing of a speed breaker and a sharp turn, will not cause the device to give out a false alarm, but a powerful blow, which is the hallmark of an accident, will make an immediate exit of the threshold.

At the same time, the GPS is always on, and it is giving the ESP32 the location latitude and longitude via a serial UART connection. The controller is always recording the location that is most up to date so in case of an accident, the system will not have to wait for making a new recording of the location. In such a way, the obtaining of the location takes only a few milliseconds after the occurrence of the accident.

When the ESP32 is certain that the MPU6050 data indicate an accident, it doesn’t hesitate to look up the most recent GPS coordinates and compose an automated emergency message. This communication consists of the accident alert text, the very first coordinates, speed if it can be gotten, the vehicle ID, and the time of the day. When the message is all set, the GSM/GPRS module is prompted by the ESP32 to send out the message to the emergency contacts that had been saved. In doing so, the device uses cellular technology to establish the connection and carry out the request. Therefore, the system can be used in areas where there is no internet.

This accident prevention plan makes the module fully capable, of course, in the most dependably way possible, to rapidly judge situations and to get in touch, then it comes in handy at those times of emergencies while at the same time being cheap and simple to deploy.

Block Diagram:

The block diagram shows the hardware architecture of a Real Time Accident Monitoring System Fig 3. Different components of the system have different functions but at the same time, they all collectively work to monitor the vehicle conditions, detect accidents, and send emergency alerts. The ESP32 microcontroller is like the central nervous system of this system which connects and manages all the parts.

Fig- 3. Block Diagram

1.      ESP32 (Microcontroller)-The ESP32 is the main hardware module which is in charge of:

•   Interfacing sensors for recording real, time data

•    Application of algorithms based on data for detection of car crashes

•    Using GSM/GPRS modules for sending and receiving messages

•   Running power management operations

•    The ESP32 is in control of the whole system field and makes sure that the right actions are taken on time.

2.       GSM Module (NEO-6M)-The intention of employing this device is to:

•    continually measure the source of the radio signals and from the different satellites obtain accurate position data (latitude and longitude)

• Tracking location through the data mentioned above becomes an essential part of the process of sending emergency alerts if there has been a crash.

3.      IMU Sensor (MPU6050)-The Inertial Measurement Unit (IMU) comprises:

•    Accelerometer, recognises the direction of acceleration (the part of the car that was hit)

•    Gyroscope, recognises the angular change (turning over of the vehicle)

•    This component detects quite sudden changes of the situation caused by a crash. The ESP32 gets the data from the IMU for accident detection.

4.      GSM/GPRS Communication Module (SIM808)-This device is what provides wireless connection to the world outside the system, thus, making it possible, among other things, for the system to:

•    The module closest to the GSM confirms and after that, it is the module that gets on with the task of Transmitting, Latitude and longitude, Time and date, Additional information (like car number, system ID)

5.           Power Module (ST6854- C)-The part of the system responsible for power management:

•    Performs voltage conversion and regulation to ensure safe operations of the ESP32 and the peripherals

Protects the system from power surges

•    If necessary, supports the implementation of charging functions

•    It is equipped to handle the normal as well as the lengthy standard of vehicle usage without interruptions.

Battery (R-41145750)-The internal battery is the one that provides the portable power for the system and guarantees that:

•    The accident alert system can still work even when the vehicle power line is cut or damaged due to the crash

•       Here is 24/7 monitoring without the need to fully depend on the vehicle power communication functions to fail.

4. Results and Discussion

The proposed Real-Time Accident Monitoring System was implemented and tested under various simulated and real driving conditions to evaluate its performance, reliability, and responsiveness. The prototype was installed inside a test vehicle, and multiple test scenarios were conducted, including normal driving, sudden braking, speed breaker crossings, sharp turns, and simulated collision events.

Accident Detection Accuracy

The MPU6050 accelerometer and gyroscope continuously monitored acceleration and orientation values. Threshold-based algorithms effectively differentiated between normal vehicle vibrations and actual collision-like impacts. During experimental testing, the system achieved an accident detection accuracy of over 80%, which aligns with similar IoT-based accident detection systems reported in previous studies. False triggers caused by potholes or speed breakers were minimized through signal verification logic and sensor fusion.

Alert Transmission Performance

Once an accident was detected and verified, emergency alerts were generated automatically and transmitted via the GSM/GPRS module. The alert messages contained precise GPS coordinates (latitude and longitude) along with time and vehicle identification details. The average alert transmission time was observed to be less than 10 seconds, ensuring rapid communication with emergency contacts and police helpline services.

System Reliability

The use of an internal battery ensured uninterrupted operation even when vehicle power was disconnected during a crash. Continuous GPS tracking enabled immediate location retrieval without delay. The system performed reliably in both urban and semi-rural test environments, demonstrating its ability to operate effectively even in areas with limited internet connectivity.

The results confirm that the proposed system significantly improves accident reporting speed and reduces dependency on bystanders or manual communication. Compared to traditional systems, the integration of ESP32, sensor fusion, and GSM communication enhances real-time responsiveness and reliability. Although vision-based systems may achieve higher precision, they require complex infrastructure and higher costs. The proposed solution offers a low-cost, deployable, and scalable alternative, particularly suitable for existing vehicles.

5. Conclusion

This paper presented a Real-Time Accident Monitoring System based on IoT technology that automatically detects vehicle accidents and sends immediate alerts to emergency contacts and authorities. The system integrates an ESP32 microcontroller with an MPU6050 accelerometer-gyroscope sensor, GPS module, and GSM/GPRS communication to provide real-time crash detection and location reporting.

Experimental evaluation demonstrated that the system achieves high detection accuracy with minimal false alarms and ensures fast emergency alert transmission, thereby reducing response time during critical situations. The proposed solution is cost-effective, reliable, and easy to install, making it suitable for widespread adoption in both new and existing vehicles.

By eliminating dependence on human intervention and enabling automated emergency notification, the system has strong potential to reduce fatalities and injuries caused by delayed medical assistance. Future enhancements may include AI-based crash classification, cloud data analytics, mobile application integration, and V2X communication to further improve accident detection accuracy and intelligent transportation safety.