safe.trAIn – Safe AI for the Next Generation of Driverless Trains
The transport sector is a major source of CO₂ emissions. Alongside increasing the share of electric vehicles, rail transport is expected to contribute significantly to achieving European climate protection targets.
Comprehensive digitalization and automation of train operations are essential for a climate-neutral and attractive transport system. Driverless trains, in particular, enhance the appeal of rail transport by enabling more frequent services (shorter cycle times), improved reliability (higher punctuality), and greater operational flexibility. Additionally, they help address the shortage of train drivers and lower operating costs, including energy and maintenance, through optimized operation.
The vision of safe.trAIn
In a safety-critical environment such as driverless rail transport, AI methods are needed that are demonstrably robust and safe, for example, in order to detect obstacles on the track with the necessary reliability.
The safe.trAIn project laid the foundations for the safe use of AI for driverless rail vehicles and thus addressed the greatest technological challenge for the introduction of driverless regional transport.
Project Results
In a consortium of technology suppliers, research institutions, and standardization and testing organizations, the safe.trAIn project team was working on combining the possibilities of artificial intelligence with the safety considerations of rail transport. The focus was on AI-based functions for object recognition.
After three years of intensive collaboration, the 16 partner companies look back on the following key results:
Security Architecture
Development of a security architecture leveraging dissimilarity and monitoring principles
Safety Verification Strategy
Establishment of a safety verification strategy for AI-driven systems
Standards and Norms
Realization of standards and norms for the use of AI with suitable ODD (Operational Design Domain) in the railway sector.
The Role of SETLabs:
Transferring Automotive AI Expertise to Rail
Building on our proven automotive expertise, SETLabs contributed with the following topics:
- AI-Based Object Recognition: Camera and LiDAR leveraging automotive expertise
- Deep Learning Sensor Fusion: Bridging perception gaps (automotive vs. rail)
- AI System Architecture: Secure perception framework for driverless trains
- Sensor Data Labelling: Training datasets for pedestrian detection in rail scenes
- Virtual Performance Testing: Evaluation of perception metrics in a VTF
With the help of AI-based functions, safe.trAIn enables driverless train operation with the aim of safe, economical and comfortable regional transport.
The project was funded by the German Ministry for Economic Affairs and Climate Action.
