Real-time train data to predict level crossing boom gate closure

iMOVE’s Using real-time train data to improve level crossing safety project, with partners Department of Transport and Planning (Victoria), Metro Trains Melbourne, and La Trobe University, has been completed, and its final report is available for download below.

Background

Informing pedestrians and road users when a train will be passing and for how long the boom gates will be closed is a key application of real-time train data to improve level crossing safety.

Level crossing safety remains a significant concern, with more than 2,500 public road and pedestrian rail crossings in Victoria alone (and approximately 20,000 Australia-wide).

Incidents at these locations can result in severe or fatal consequences, underscoring the urgent need for innovative safety measures.

People with disabilities often experience stress when navigating level crossings, due to fears of tripping, becoming trapped between closing gates, or not reaching the other side in time.

Providing accurate, real-time crossing information can make their journeys significantly safer and less stressful.

Background

To address long-standing challenges around level crossings, the Department of Transport and Planning and the Metro Trains and Victorian Railway Crossing Safety Steering Committee’s Rail Crossing Human Factors Group facilitated a series of workshops.

These workshops explored technological solutions and identified practical use cases to enhance level crossing safety.

A key outcome was a focus on leveraging real-time transport data to support vulnerable road users, particularly people with disabilities, by providing timely and predictive safety information.

The goal is to provide advanced warnings, including clear estimates of how much time remains before boom gates close, thereby reducing anxiety and improving confidence for users who may require more time to cross safely.

At the centre of this project is the development of an algorithm to estimate level crossing closure times based on rail time train position from DTP open data.

Best practices

Inclusive design principles, risk-based assessments and varying levels of crossing controls guide Australia’s national best practices.

Strategic initiatives like the Victorian Technology Readiness Roadmap and the emerging Australian National Access Point platform also aim to future-proof mobility data sharing by supporting interoperability and long-term adoption across jurisdictions.

Internationally, initiatives such as the Netherlands’ TNO Crossing Aid and France’s C-ITS-integrated smart crossings led by SNCF demonstrate how real-time data, connected vehicle technologies and user-focused design can reduce risk, enhance confidence and support vulnerable road users.

End user expectations

The project interviewed people with disabilities at 11 pedestrian level crossings in Greater Melbourne to understand their behaviours, challenges and preferences for receiving time-to-cross information.

Many feared tripping, narrow paths or alarms triggering mid-crossing. Most welcomed the idea of a countdown timer with traffic light cues to reduce anxiety.

Desired warning time ranged from 5 to 30 seconds for experienced users, to 2 to 3 minutes for more cautious individuals.

Users favoured in-situ solutions with visual, audio and haptic feedback.

Machine learning prediction algorithms

To predict boom gate closure times in a field trial at Diggers Rest Station, the system – dubbed CrossSafe – applied machine learning algorithms trained on real-time train data to estimate train arrival times.

By combining infrastructure data with real-time train information – drawn from Metro, V/Line and 4Tel – it achieved sufficient accuracy in predicting boom gate closure events.

La Trobe University’s Centre for Technology Infusion designed and deployed the real-time level crossing alert system at Diggers Rest Station.

To capture the non-linear and dynamic patterns of train movement, a Random Forest Regressor was selected as the prediction model. This ensemble method is well-suited for regression tasks with complex interactions among variables.

To improve prediction accuracy, it integrated trip information and timetable data into the model.

These inputs help identify specific train runs, estimate scheduled arrival times and determine travel directions, providing a richer context for decision-making. With these enhancements, the predictive model becomes more reliable and context-aware.

C-ITS compliant prototype development

Existing European Telecommunications Standards Institute (ETSI) ITS-G5 Release 2 specifications – particularly Decentralised Environmental Notification Messages – can effectively support VRU protection without requiring additional roadside infrastructure.

By leveraging 4G networks, cloud-based processing and real-time predictive algorithms, the system delivers timely alerts to pedestrians and cyclists based on actual train movements.

This approach enables a scalable, efficient and future-ready solution, laying the foundation for integrating vulnerable road users safety at level crossings into ETSI-compliant C-ITS ecosystems and addressing the unique challenges of road–rail interactions.

To broaden the system’s impact, future enhancements will focus on smarter algorithms, accessible interfaces and integration with public ITS platforms such as Google Maps and TomTom.

Report findings

Grounded in real-world validation, the prototype successfully demonstrated real-time alerting, machine-learning–based ETA predictions and inclusive mobile interfaces for VRUs, particularly disabled and elderly pedestrians.

Field trials at Diggers Rest confirmed the system’s practical feasibility, delivering high prediction accuracy (average errors ranging from 5 to 16 seconds), responsive warnings and positive user feedback. These results validate not only the system’s technical merit but also its user acceptance and operational readiness.

Globally, similar deployments in countries such as the Netherlands and France reinforce both the need and the viability of this approach.
However, this solution distinguishes itself with its predictive intelligence, real-time data integration and adherence to C-ITS standards. This ensures compatibility with emerging mobility ecosystems, from connected vehicles to assistive wearables.

Conclusions

This project is a stepping stone to further opportunities for level crossing enhancement. Leveraging its framework and consolidated algorithms, the approach is scalable and adaptable, enabling future deployment across all Victorian stations and potentially nationwide.

By integrating crossing closure predictions into journey planning, it can also enhance travel efficiency and safety, potentially extending to predictive capabilities for passive crossings to address safety gaps where there are no active signals.

To progress CrossSafe to its next stage, a follow-up project is proposed, focused on:

• Securing support from transport agencies in other states
• Integrating with required data sources in each jurisdiction
• Partnering with relevant third-party apps
• Maturing and end-user validation of the solution

Expected project impacts

This project is just one of the many case examples where data can help improve accessibility and safety.

It is addressing an important issue, and one can only eat an elephant bite-by-bite, but we hope that by further development of Boomtime, we can help pave the way for more real-time data-driven accessibility solutions..

Erik van Vulpen, Deputy Director Centre for Technology Infusion at La Trobe University

Download the final report

Download your copy of the final report, Using real-time train data to predict level crossing boom gate closure, by clicking the button below.

DOWNLOAD THE REPORT