Car2x and AI: Intelligent Technologies for Improved Safety in Road Tunnels
How can the combination of artificial intelligence (AI) and Car2x help to make road traffic safer? How can dangerous situations in road tunnels be detected earlier and thus mitigated? The project "Künstliche Intelligenz zur Verbesserung der Sicherheit von Tunneln und Tunnelleitzentralen" (KITT for short) dealt with precisely these questions and was successfully concluded with a demo in the Rosenstein Tunnel in Stuttgart, Germany.
?
AI-Based Tunnel Monitoring
Highways and federal roads with their tunnels and bridges not only connect regions within a country, but also countries with each other. To further increase road safety in road tunnels, Germany and Austria initiated the KITT project.
The aim of the project was to develop a tunnel monitoring system that uses AI-based data processing to analyze various data and evaluate it regarding potential risks. The data comes from sensors inside the tunnel and from vehicle information sent by radio to the Car2x receiver units in the tunnel.
The AI then recognizes a wrong-way driver, for example, based on the CAM messages (Cooperative Awareness Message) sent by radio from the vehicles. Among other things, this contains the current position, orientation, speed, and other internal vehicle information. The AI can use this information to analyze the vehicle's behavior and determine whether it is driving in the wrong direction, and then issue an alarm accordingly.
The figure above shows the various elements that play a role in tunnel monitoring, from data acquisition to hazard warning.
Testing in the Rosenstein Tunnel in Stuttgart
On December 7, 2023, the KITT system was tested with a demonstration in the Rosenstein Tunnel in Stuttgart (Germany), as the tunnel was closed at night due to maintenance work. The scenarios supported by the AI system include a variety of dangerous driving situations. For this reason, Vector's CANoe.Car2x simulation and test tool were used, which make it easy to simulate scenarios without real vehicles. The scenarios include wrong-way drivers, accident situations, stationary vehicles and slow vehicles. These were created using the CANoe scenario editor based on the requirements, and each contains around 10 vehicles to cover both normal traffic flow and more abnormal situations. Each simulated vehicle generates its own CAM messages, which were sent via a VN4610 measurement interface for sending and receiving 802.11p frames. The VN4610 was located in the tunnel tube and was connected via the tunnel Ethernet network to the CANoe PC, that was located in the tunnel control room. In the middle of the tunnel, a roadside unit (RSU) was installed in the north tube in advance. The distance between the RSU and the VN4610 was approx. 400 m. The RSU receives the CAM messages sent and forwards them to the KITT system. Based on the content, the AI module recognizes potentially dangerous situations and displays them on a graphical user interface (see video from SPIE OSMO and Vector), which warns the tunnel control center immediately. During testing, the simulated traffic situations in CANoe and on the graphical user interface of the KITT system could be observed in real time in the control room. This allowed the overall system with all the necessary components such as the KITT system and the installed RSU to be finally verified.
From Prototype To Real System: What's Next?
The KITT project has shown how versatile Car2x can be. It combines many different subject areas, some of which have not yet worked together in this form, to make our roads safer.
As the KITT system is a prototype that has so far been tested in the laboratory, the follow-up project KITT-Pro is planned. Here, the KITT system is to be tested in a real environment over a longer period of time, including with real vehicles. The topic of positioning in tunnels will also play an important role here.
?