
Case study
Improving the operational performance of the London Underground
We collaborated with a global engineering, architecture and design company to improve the operational performance of the London Underground.
Customer
Arup is a global leader in design, planning, engineering and architecture, working across every aspect of the built environment. Arup’s City Modelling experts spotted an opportunity to create an interactive digital representation – a type of digital twin – using existing data, allowing them to uncover new insights about the performance of the London Underground.
Solution
The data engineering challenge started with two years of historical data from three Transport for London open API feeds – timetables, live line status reports and arrivals data – in widely varying formats and quality. We identified, extracted and transformed the key data and used it to populate a graph representation of the network.
We then used the model to explore and understand the performance of the London Underground using data science techniques. The tool reveals opportunities to improve operational performance and customer experience with targeted investment. Using machine learning, the model also allows Arup to analyse the causes of delays, where delays tend to occur, and how delays spread across the network.
Impact
The Network Performance Tool has been deployed on Arup’s infrastructure; you can see a visualisation of the insights generated by the model on the variation in travel times on three typical journeys here.
Early predictive experiments have already highlighted information which could significantly improve performance; for example, experiments suggest that delays on one line can have knock-on effects on other lines with a shared station. With this information, network controllers could perform real-time interventions to minimise delays and further improve services and customer experience.