On 7 May, at the UIC Headquarters in Paris, the first Passenger Services Group Innovation Workshop planned for 2024 took place: “Transforming the Passenger Experience with Generative AI”.
Vittorio Carta, Chair of the UIC Passenger Services Group, opened the Innovation Workshop and welcomed the participants, attending either in person or online.
Participants represented key rail industry players, including SNCF, Trenitalia, FSTechnology, SBB, FGC, MÁV-Start, LTG Link, CP, Entur, OEBB, CD, SZ and UIC.
Following a quick round of self-introductions, Stefano Scarci (EY) provided an overview of Generative AI and of AI impact on the travel industry. While traditional applications of Machine Learning are aimed at implementing individual siloed models that require task-specific training, Generative AI is largely based on massive, pre-trained, generalised Foundation Models. Such models can however be adapted to specific tasks via fine-tuning or Retrieval Augmented Generation. Practical examples were shown during the session, including a Q&A on UIC technical documents, automated travel inspiration copy creation based on passenger profile, and AI-powered travel planning. The overview also covered the main risks and challenges relating to Generative AI, its emerging applications in passenger rail and the results of a recent study performed by EY, involving over 5000 subjects in five European countries, on the propensity of travellers to interact with Generative AI when planning, booking and receiving assistance when travelling.
Adèle Huber and Eva Krueger from SBB gave a presentation titled “On track with AI: A glimpse into SBB’s GenAI Initiative”. SBB has defined a comprehensive AI Initiative that encompasses five main fields of action: recognising digital responsibility; realising business opportunities; mastering change in the world of work; developing AI competences; and providing AI service. Al literacy, in addition to business and digital skills, is important for all managers at SBB. SBB intends to improve the customer journey with AI. However, it realises that customer interactions differ in complexity and emotionality: emotional customer needs are often better addressed by human attention. Therefore, digital interactions are evaluated differently depending on the customer group.
Claire Nicodeme from SNCF gave a presentation entitled: “GenAI: Improving Data Augmentation for better training”. Data Augmentation is the process of artificially generating new data from existing data, primarily to train Machine Learning models. It helps to enrich training datasets by creating many variations of existing data and improves the overall performance of ML. It reduces the effort required to collect and prepare data, mitigates overfitting and improves privacy. This helps to overcome the typical issues encountered in rail use cases. For example, models aimed at improving rail maintenance by predicting defects are often biased because of the scarcity of defects/failure data (luckily, defects are uncommon). In passenger flow analysis, GDPR strictly restricts the use of personal and biometric data, which limits the ability to develop, test and validate ML models. Generative AI can create the required data.
Lorenzo Ferrone and Veronica Fabiani from FSTechnology, the tech company in FS Group, presented “UNTIE: Exploring new approaches for tackling Railway Traffic Optimisation”. Railway Traffic Optimisation includes Real Time Routing and Optimised Platforming, i.e. how to assign platforms so that users’ walking distances are minimised. Promising results have been obtained in FSTechnology by applying Quantum Computing and Reinforcement Learning. The UNTIE platform is a patent-pending concept proposed by FSTechnology that includes a simulation module, consisting of a railway digital twin that implements the operating rules, an optimisation module that includes a quantum computing and a reinforcement learning environment, and a visualisation module that supports decisions made by rail agents.
In the breakout sessions, participants had the opportunity to discuss and co-create ideas about three key topics for the rail industry related to Generative AI:
- In what ways could railways apply Generative AI to enhance the passenger experience and customer service?
- What are the potential uses of Generative AI in railway operations to assist employees and advance operational effectiveness?
- How can UIC facilitate international initiatives to orient and support railways on their journey to implementing Generative AI?
During the final plenary session, the outcomes of the breakout sessions were presented.
