Artificial intelligence for railroads: Potential and challenges for application

Abstract

It is clear that AI methods have a great deal to contribute to the solution of relevant problems. Realistic expectation management for AI solutions is important for a productive compromise between the extreme “AI can solve everything” and “AI is unusable” positions. AI tools should therefore not be considered in isolation, but approached as an effective tool that can harmonize with other methodologies. This secures a connection to established procedures and enhances acceptance by stakeholders. It is also clear that there is a need of guidelines and procedures for approving AI algorithms, especially in safety-critical applications. Since the capabilities and reliability of AI methods will continue to grow in the future, it is important to start preparing for their use by initiating suitable approval procedures.