What it does
The system detects abnormal vehicle stay time in tunnels using Hi-Pass (RFID) data. If no exit is detected, it checks for accidents via temperature, smoke sensors and radar, then alerts incoming vehicles via LEDs and displays to prevent secondary collisions.
Your inspiration
According to the Korea Expressway Corporation, secondary accidents in tunnels increased from 2.7% in 2019 to 3.9% in 2024 (as of August), with an average fatality rate of 55.9%—5.7 times higher than all tunnel accidents (9.8%). In addition, 20 highway patrol accidents occurred from 2019 to 2023, causing four deaths (Ministry of Land, Infrastructure and Transport, 2024). These statistics highlight the severe risk of secondary accidents, especially in short tunnels under 500m where detection systems are lacking. Recognizing this, we developed a low-cost, real-time accident detection and warning system using existing Hi-Pass infrastructure.
How it works
The Arduino-based system runs in five steps: ① initialization ② vehicle detection ③ accident check ④ alert output ⑤ reset. ① Initialization All modules (RFID, sensors, LEDs, servomotor) are activated and set to ready. ② Vehicle Detection RFID readers at the tunnel's entry and exit log each vehicle's UID and entry time. Excessive stay time triggers the accident check. ③ Accident Check Temperature and gas sensors determine the case: high readings signal fire; otherwise, it's congestion. The system displays a message and adjusts LED/buzzer accordingly. ④ Alert Output LEDs and NeoPixels give visual warnings. For fires, red LEDs light and the buzzer sounds. The servomotor rotates an ultrasonic sensor to find obstacles and lights the matching NeoPixel spot. ⑤ Reset If the vehicle exits normally, records and alerts reset for the next detection. This low-cost system enables real-time detection, classification, and visual alerts using existing Hi-Pass infrastructure.
Design process
First, after analyzing domestic tunnel traffic accident statistics and related laws and regulations, it was confirmed that short-distance tunnels under 500m are not subject to mandatory detection devices. As a result, we started development under the judgment that a system that can supplement the detection blind spots was needed. The high-pass infrastructure was considered usable during planning. The fact that most vehicles are already equipped with RFID tags, which allows the analysis of the entry and exit times without the need to install additional equipment, was also a major advantage in terms of economy. Since then, the experiment has been conducted with RFID readers and tags based on Arduino so that conditions similar to high-pass data can be simulated. If the vehicle does not recognize the exit tag for a certain period of time even after entering the tunnel, the smoke and temperature sensors are activated through ESP32 to check for congestion or fire. It has also implemented a function to display the location of the accident through an LED display and to give a visual warning to outside vehicles. Finally, a sensor module housing was manufactured using a 3D printer and tested in a model environment that simulates an actual tunnel to ensure operational stability.
How it is different
Unlike existing expensive sensor-based systems, this one reduces installation costs by using existing high-pass infrastructure. High-pass data is used not only for tolling but also creatively reinterpreted for accident detection. Most tunnel safety systems focus on medium and long-range tunnels, leaving short tunnels as blind spots. Our system addresses this with lightweight sensors and microcontrollers, overcoming spatial limits. Similar systems exist, but most rely on fixed CCTV or manual detection, whereas ours automatically analyzes stay times—making it rare and distinct. Server-based log storage also enables data-driven analysis and policy support beyond simple detection.
Future plans
In the future, we plan to expand the platform so that detected accident information can be visualized in real time and transmitted to smartphone apps or traffic control centers. In addition, we plan to implement a function to predict the likelihood of accidents by introducing machine learning analysis based on vehicle traffic logs and accident data stored in servers. The goal is to work with real local governments and public institutions to facilitate pilot installations, verify the performance of the system on site, and spread them to short-haul tunnels across the country.
Share this page on