Automated Train: Revolutionizing the Railroad

The progressive implementation of automated trains marks the start of a revolution in German rail transport which is set to have far-reaching effects on efficiency and sustainability in the field of private and commercial logistics. This is because there are various levels of automation for trains, such as fully automated trains without drivers and without staff responsible for checking the train’s doors on departure and evacuating it in an emergency. Approaches which aim to achieve the full automation of the railways without AI are usually more costly, however, as they require a stronger infrastructure. For this reason, Germany is positioning itself as a global pioneer by pursuing the development and implementation of automated trains with the support of AI. 

After all, automated trains will contribute to the modernization of the transport system and strengthen Germany’s position as an innovation leader in the global rail transport industry. Nevertheless, there is a need to clarify how AI controlled processes can be reconciled with the requirements and authorization procedures for driverless trains. This is where the safe.trAIn project steps in. 

The safe.trAIn project has the goal of laying the foundations for the safe use of AI for driverless regional trains. The aim is to bring fully automated trains, which have to date only operated in closed and controlled environments, to more open environments such as regional rail transport, where it will also be necessary for obstacles such as people or fallen trees to be detected reliably by the automated trains. 

Safe.trAIn is developing test standards and methods for the use of self-driving trains in the automation of rail transport. The project is demonstrating how these standards can be applied to the specific use case of the driverless regional train. The focal points of the project are: 

• AI-based methods for automated trains 
• Safety certificates for trains with AI components 
• Test procedures and methods for the verification 

The project consortium consists of 17 partners, including businesses from the rail industry, technology providers, research institutions as well as standardization and testing organizations. The project has received funding from the Federal Ministry for Economic Affairs and Climate Protection in total of approximately 23 million euros. 

Within the framework of the safe.trAIn project, Fraunhofer IKS is responsible for developing methods and tools to prove the trustworthiness of AI methods. Automated trains should be designed to be as safe and efficient as possible, which can primarily be achieved through a traceable explainability, robustness and fulfilment of safety requirements of the associated AI systems. Fraunhofer IKS is focusing on three key topics: 

• The description and monitoring of the operational design domain (ODD) for rail transport 
• The robustness of both AI and monitoring mechanisms for AI-based systems 
• Proof of safety for AI functions in terms of the train’s ability to perceive its surroundings 

Fraunhofer IKS is also working on processes for the quality assurance of the AI functions during the developmental stage. In this respect, one focal point is on conducting research into the methods that increase the transparency and interpretability of AI applications for automated trains. 

The integration of automated trains into rail transport in Germany appears likely to have a transformative impact which will improve the safety, efficiency and sustainability of the transport system. Fraunhofer IKS is focusing on two work packages in particular: 

• Proof of safety for AI methods: Fraunhofer IKS is responsible for the development of test methods and tools to verify the trustworthiness of the AI methods. Methods are being researched to ensure the explainability, robustness and safety of the AI applications. 
• Safety architecture for AI-based functions: Fraunhofer IKS is researching the safety architecture for AI-based functions in fully automated operations (GoA4). In this respect, particular attention is being paid to the definition and influence of the operational design domain to allow safe operation of the automated trains in the defined environment.

Fraunhofer IKS is also working on standardizing the results of the project. For example, DIN DKE SPEC 99002 “Terminology – AI in railway applications” is being developed, as well as other standards. 

The safe.trAIn project is making an important contribution bringing rail transport into the future. It is laying down the foundations for further automation solutions and providing important additional support to the current standardization activities.