How do you find a road that isn't on the map?

In the fast-moving world of urban development, road networks are constantly changing—new streets appear, old routes close and construction projects reshape the way we navigate. For cities, keeping these maps accurate isn’t just a matter of convenience—it’s essential for safety, efficiency and smart planning.

That’s why HERE Technologies and RMIT University have teamed up to develop an innovative approach to spotting “missing” and “blocked” road links using a surprisingly simple—yet remarkably effective—analytical technique: the shortcut method.

Tackling a global challenge

Road network maintenance is a daunting task. Traditional methods often lean heavily on statistical models or machine learning (ML) algorithms, which work well under ideal conditions—but can falter when data is scarce or irregular.

The HERE–RMIT team set out to explore a new path: a network-analytical approach that doesn’t just rely on volume or frequency of data, but on the relationships between how people actually move and how the map says they should move.

To get there, they evaluated an algorithm using road network and trajectory data. At a high level, the experiments included calibration, performance, ablation and blocked links—leading up to what the team calls the shortcut method:

• Calibration: RMIT tuned algorithm parameters to find optimal settings and tested accuracy without altering the road network.

• Performance comparison: the algorithm was benchmarked against a recent machine learning–based method to evaluate relative effectiveness.

• Ablation study: the team examined how the length of missing road links (shorter vs longer) impacts performance.

• Blocked links: the algorithm was applied to real-world cases to detect blocked roads in Eindhoven, Netherlands.

The shortcut method compares the real-world travel paths captured in vehicle trajectory data with the shortest routes predicted by the map. Significant discrepancies between the two can reveal a missing link (a road that exists but isn’t in the map) or a blocked link (a road in the map that’s inaccessible in reality).

Why the shortcut method stands out

• Performs in tough conditions: it matches, and often exceeds, the performance of machine learning models, especially when GPS data is sparse or sampled at low rates.

• Dual functionality: detects both missing and blocked links using the same logic.

• Interpretable results: the method is transparent, making it easier for planners and engineers to trust and act on its findings.

• Scalable to data-limited regions: because it doesn’t depend on high-quality, high-frequency probe data, the shortcut method is suitable for cities where such data is rare.

Putting the method to the test

The team validated the shortcut method using multiple real-world datasets, including HERE’s commercial probe data for Eindhoven, and publicly available data from Chengdu, China and Rome, Italy.

Their key results included:

• A higher recall than ML methods in identifying missing links, particularly shorter ones that are notoriously tricky to detect.

• The successful detection of real-world blocked links around the PSV Eindhoven stadium during construction closures.

• The ability to infer topological changes in Rome’s road network over time by comparing historical and modern maps.

A step forward for traffic management

The HERE–RMIT collaboration offers city planners, transport authorities, and mapmakers a powerful new tool. By detecting inconsistencies early and reliably, cities can update their maps faster, inform navigation services more accurately, and ultimately keep people moving efficiently.

Data in action without compromising privacy

While HERE’s detailed road network and trajectory data powered the Eindhoven experiments, the collaboration ensured data security. No raw trajectory data was shared publicly; only high-level summaries and anonymized visuals were included in the research.