An AI-monitoring system that can track passenger numbers on trains to prevent overcrowding has been developed by Loughborough University computer scientists.

The technology uses depth-sensing cameras and onboard AI to monitor passenger numbers and movement throughout the network. Rather than recording conventional video footage, it captures depth information only, allowing passenger flow to be monitored without identifying individuals.

The researchers, who worked with rail technology company TrainFX on the technology, said the system could help operators better manage capacity, improve passenger information and reduce overcrowding across the network.

Currently, operators often rely on delayed or incomplete data, meaning some carriages become overcrowded while others still have space available. The AI system is capable of estimating how crowded each train carriage is in real-time, even during busy rush-hour periods and in low-light conditions.

“This collaboration shows how responsible AI can support the future of transport,” said Professor Baihua Li, project lead. 

“By processing passenger flow data on board and using privacy-conscious depth-imaging technology, the system can provide real-time carriage occupancy insights to operators and passengers.

“Our aim is to develop AI that is not only technically robust, but also trustworthy, practical and centred on improving the passenger experience.”

Embedded directly within TrainFX’s Smart Passenger Information System, the technology allows data to be processed directly on board trains and live occupancy information to be shared with operators and station staff.

The technology could also help operators to improve scheduling, crowd management and longer-term service planning. In the future, the data could also be shared with passengers, allowing them to identify less crowded carriages before boarding.

The multi-camera prototype has already been installed and tested within TrainFX’s simulated environment, where it monitors passenger movement around train doors and carriages. The researchers said the AI model has shown high levels of accuracy and reliability during testing.

The prototype is now ready for live train trials, with TrainFX working alongside train operators on the next stage of deployment in public railway environments.

Hansoon Han, managing director at TrainFX, said: “Overcrowding is one of the biggest frustrations for passengers, especially when some parts of the train are much busier than others.

“This project is about giving operators better real-time information, so they can manage daily services more effectively and build a clearer picture of passenger demand over time.”