Have you ever been stuck in traffic, rerouted through endless detours, all because of multiple roadworks happening in the same area? Why do cities – equipped with smart data, digital maps and planning tools – still struggle to coordinate maintenance in a way that does not paralyze mobility? This blog shows how Aimsun, in collaboration with the University of Deusto, developed a simulation-based AI-driven optimization to strategically coordinate and schedule roadworks in the South Holland Use Case of the EU-funded SYNCHROMODE project.
The Challenge: Coordinating complex, long-term roadworks
South Holland (The Netherlands) is preparing for extensive, long-term roadworks essential to maintaining and upgrading its infrastructure. However, closing multiple roads simultaneously forces drivers onto longer alternative routes, increasing congestion and delays. Planning still relies heavily on expert intuition, often resulting in inefficient coordination. While the MELVIN platform – the Dutch digital register for roadworks – provides a valuable platform to support coordination between municipalities, it is not yet fully leveraged for predictive, optimized planning across projects.
The solution: simulation meets AI-driven optimization
To tackle this, we present the following framework that integrates three core components (see Figure 1):
- A data preparation module that extracts, filters, and process real-world roadwork data from MELVIN into structure capacity reduction events.
- A transport simulation module that evaluates the network performance under each scenario using a multi-day traffic assignment methodology.
- An AI-based optimization module that iteratively refines candidate schedules based on simulation outputs.
Figure 1. Flowchart of the core components of the roadworks planning module
These components are connected via an automated Python-based workflow that interfaces with the Aimsun Next simulation environment and manages the transformation and evaluation of scheduling scenarios.
Each candidate’s schedule is encoded as a set of flexible start dates for the roadworks and is translated into a simulation configuration. The simulation model estimates the impact of each scheduling strategy on the transport network performance over time, accounting for both full closures and speed reductions in affected sections as well as their cumulative effects on general traffic and Public Transport (PT) services. These outputs generate KPIs that feed the optimization module, which explores alternative roadwork plans to minimize the overall disruption impacts.
Case Study: The Hague (South Holland) in Focus
This framework is applied to the northern part of The Hague, to help transport authorities identify the ideal start times, prioritize projects, and coordinate mitigation strategies such as rerouting and public communication.
The simulation models a road network with over 15,000 sections and 7,000 nodes, covering traffic demand between 8-9 AM based on a calibrated 2018 weekday pattern. It evaluates different start dates for 20 roadworks projects over an 18-week planning horizon (January 1 to May 1, 2025). Each project can begin in any week, and once started, may span multiple weeks.
Figure 2. On the left, part of the South Holland region. On the right, subnetwork of the northern part of The Hague in Aimsun Next. In red, full closures. In blue, reduced speed limits.
Key modelling assumptions include:
- Fixed durations and weekly time resolution for roadworks, meaning that changes in road capacity can only occur between weeks, not within individual days.
- Capacity restoration after project completion.
- Static demand profileduring the peak period, with no modal shift.
- Macroscopic network loading and static traffic assignment.
Each scenario is assessed using two Key Performance Indicators (KPI):
- Total travel time across the network, as a network-wide measure of traffic conditions.
- Number of cancelled public transport lines, as an indicator of the impact on PT services.
The optimization module uses a Bayesian Optimization to efficiently explore the scheduling space, guided by a Gaussian process surrogate model and an acquisition function. A scalar objective function drives the optimization process by combining the two KPIs into a single metric.
Results from the application
Figure 3 compares the original MELVIN-based roadwork schedule (on the left) with the optimized schedule produced through proposed methodology (on the right). The original schedule had significant overlaps of major roadworks causing peak disruptions on certain weeks. The optimized schedule distributes these activities more evenly over time, reducing conflicts and minimizing disruptions.
Figure 3. Gantt chart comparison between baselines and optimized schedules
Figure 4 illustrates the weekly evolution of the KPIs. Each point on the graph represents the KPI measured during a specific simulation week. The optimized roadwork schedule reduces total travel time and PT cancellations compared to the original plan. It smooths disruptions over time, especially improving performance during weeks 5-6 and 14-16, bringing results closer to the ideal baseline.
Figure 4. Weekly comparison of total travel time (top) and cancelled PT lines (bottom) for original versus optimized roadwork schedules.
The following clip provides a visual analysis of the V/C ratio under the optimized roadwork schedule over the simulated weeks. Most roads show low to moderate traffic levels, with only a few areas experiencing congestion. The spatial impact analysis confirms that increases in traffic congestion are mostly minor and localized near work zones, demonstrating the effectiveness of the optimization in minimizing network-wide disruption.
Figure 5. Colored map, V/C ratio optimized solution
Why it matters? – Benefits
The implementation of this roadworks planning module allows transport authorities and operators:
- Enhance coordination between municipalities. Shared data and optimized timelines reduce conflicts between overlapping projects. By strategically sequencing roadworks, it helps restore normal traffic conditions more quickly.
- Reduce network disruption and speed up recovery. Optimized scheduling minimizes congestion and delays, even with concurrent projects.
- Improved decision-making. Authorities can move beyond intuition, using simulation and AI optimization.
- Scalable and adaptable framework. The approach can be extended to other regions and adapted to different infrastructure contexts.
Looking ahead
Future developments will focus on refining the framework to better reflect real-world needs. This includes introducing new KPIs that give greater weight to the impact of roadworks on public transport services, recognizing its role in mitigating disruption. The framework will also be tested over a larger geographic area to assess its scalability. A key next step is piloting the prototype in collaboration with transport authorities in South Holland, bringing the tool closer to real-life use.
Stay tuned for more regular updates on our LinkedIn page SYNCHROMODE project.
Acknowledgements
The present research was carried out within the research project “SYNCHROMODE”, which has received funding from the European Union’s Horizon Europe research and innovation program under grant agreement No 101104171. Additionally, we would like to thank the region of Holland Rijnland for providing their regional transport model to accomplish this study and the MAPtm team for providing very useful insights and leading this Case Study.
