Melbourne network efficiency: Improving with AI and digital twin

The aim of the now completed AI optimisation for transport corridor performance project was to leverage AI algorithms and machine learning to optimise the performance of motorways, arterials, and ramps in cities, by predicting future traffic speed and proactively responding to them. The final report from the project is available for download below.

This research aimed to revolutionise transport network management by validating an innovative virtual AI-driven Traffic Management Centre (TMC). The core vision was to harness the predictive power of Artificial Intelligence (AI), Machine Learning (ML), and high-fidelity digital twin simulations to move from a reactive to a proactive paradigm of traffic control.

The ultimate goal was enhancing network efficiency, reducing travel times, and improving the movement of people and goods across Melbourne.

Participants on the project with iMOVE were Cubic Transportation Systems and the University of Melbourne.

Methodology

Data for this research was constructed from two high-volume data streams:

1. Sydney Coordinated Adaptive Traffic System (SCATS), providing granular traffic volume, detector counts, and signal phase information; and
2. TomTom Live Speed data, real-time insights into traffic flow and speed dynamics across the network.

These two data streams were synthesised and used to create a high-fidelity digital twin of a key Melbourne road network.

The digital twin comprised 28 signalised, minor and major Melbourne intersections, all with high traffic volumes.

Across those intersections 36 distinct traffic scenarios were comprehensively analysed, exploring the network-wide impacts of interventions ranging from signal timing modifications and speed limit adjustments to full and partial road closures.

Conclusion

This research achieved it technical objectives and laid the essential groundwork to offer Melbourne a clear path forward to effectively manage the complexities of the city’s urban growth and mobility needs.

“A wide array of findings emerged from the analysis, leading to several key insights, such as the importance of tailoring control strategies to specific network segments and time periods, the benefits of implementing Adaptive Signal Timing, and the value of large-scale digital twins in supporting data-driven traffic management decisions.

From the AI forecasting model perspective, the results underscore Strategy 2’s robustness and adaptability across varied scenarios. Moreover, forecasting accuracy appears to be more strongly influenced by underlying traffic conditions than by network interventions alone.

It is also worth noting that the AI model demonstrates efficient inference, with each prediction taking about 70 microseconds, which makes it well-suited for real-time decision support, enabling timely and effective traffic management.”

Download the final report

Download your copy of the final report, AI Optimisation for Corridor Performance, by clicking the button below.