Railway performance assessment from motion analysis of mobile data: a machine learning approach

Passenger comfort and safety are critical for rail operations. Although there are some commonly used standards or methods for evaluating ride comfort, they only apply to passengers in good health and seated in stationary positions. There are few standards that address the ride safety of passengers in rougher ride conditions, and not every train journey is measured or monitored. To fully characterise passenger comfort and safety, it is necessary to consider not only the motion environment but factors such as acoustic noise, temperature, pressure and standing position.

This project develops tools and algorithms to allow passengers to measure ride comfort and safety themselves using their smartphone or other wearable devices. It also aims to collect crowdsourced data in order to allow railway companies to continuously monitor its railway network and receive instant feedback from passengers regarding the ride quality on trains.

This project will further pursue developments in efficient deep learning for resource constrained devices.