Challenge: Use AI to improve methods for providing accurate estimates of passenger loads in real time, analyzing user trajectories and gauging their response to signage

Solutions: An AI-driven system for the exploitation of multi-source data for load prediction and analysis of user trajectories

Benefits:

• Improved user experience
• Better communication
• Heightened passenger safety
• Improved line operations
• Better staff responsiveness

LINCOLN was asked to help in addressing the needs of the metro system in one of Europe’s most populated cities, in particular by predicting passenger numbers in real time, and by analyzing user trajectories as well as user response to signage in stations.

Prediction makes it possible to alert passengers as to when loads are likely to be heaviest, allowing them to adapt their behavior and thereby offering a more pleasant, safer passenger experience; it also enables agents to react more quickly to unforeseen events. Providing accurate estimates in real time calls for cross-referencing data from a variety of sources, e.g., weighing sensors, tele-ticket validations, crowdsourcing and network topology.

Analyzing how users interact with the space in metro stations is also essential for enhancing their experience and guaranteeing their safety. This can be achieved by studying data on user trajectories collected by video surveillance cameras in stations, in strict compliance with regulations on personal data. Understanding how users react to the various signs in the station is another important factor.

To achieve this, LINCOLN developed deep learning algorithms using computer vision tools such as YoLo and DeepSort, designed to detect and track users in space. A number of challenges, including the quality of the video images, the presence of obstacles (such as poles) and the massive flows of passengers at peak times, complicate individual tracking. The solution developed enables the network manager to optimize the use of space, identifying the busiest and potentially riskiest areas. Important insights linked to insufficient or poorly designed signage were gained by comparing stations with and without signage, and by observing user responses to the signage.

The experience demonstrated the significant benefits of data and artificial intelligence for improving urban rail transport, ranging from streamlining operations and heightening the passenger experience, to predictive maintenance, pricing, and traffic optimization. The use of data coupled with AI technologies is an essential lever for meeting the challenges of mobility and transforming urban transport networks into intelligent, responsive systems adapted to user needs. By combining high precision monitoring, advanced forecasting techniques, automated detection systems, and sophisticated deep learning models – in strict compliance with regulations on personal data – modern metros can significantly enhance safety and efficiency, and the overall passenger experience.