The project seeks to enhance the management of vast data volumes in modern railway transport systems. Presently, numerous sensors collect large datasets that are costly and prone to errors when processed. Our initiative will develop an innovative data-management system employing advanced Large Language Models (LLMs). These models represent significant AI breakthroughs over the past decade and will be foundational in creating Smart Assistants. These Smart Assistants are autonomous agents designed to aid human technicians in effectively processing critical data captured by sensors. This includes identifying issues such as sensor malfunctions and maintenance requirements. A key innovation in our approach involves integrating rigorous statistical methods to validate the reliability of the LLM outputs. This means that the autonomous agents will provide input only when there is high statistical confidence in their accuracy, significantly reducing the likelihood of errors during operations. The project not only aims to streamline data handling, thereby reducing operational costs and improving accuracy, but also to potentially establish a new standard for data processing within the railway transport sector. By blending cutting-edge AI with rigorous statistical principles, each agent will be finely tuned to perform its designated tasks reliably, introducing a transformative edge to data management practices in the railway industry.

This project will deliver a Machine Learning powered simulation tool called DataSim, to empower rail operators with the ability to explore the impact of timetable changes on the network state. With DataSim, operators could simulate different scenarios and find the optimal solutions for efficient and reliable rail scheduling. DataSim uses a map interface which intuitively depicts rail asset position. Through this, rail operators can view the differences between planned and actual train positions, analyse the time lost or gained at each section of the journey, and view the arrivals and departures for each station on the network. This allows rail operators and decision makers to take a scientific approach to data analysis, mitigating management problems by empowering analysts to visualise and simulate services and the cascading impacts of unexpected delay scenarios including weather, trespass incidents, hardware failure, and other unforeseen circumstances.