A European research project built to answer a deceptively simple question has concluded. The question: can a city work out how safe a street will be before it builds it?
The PHOEBE project, short for Predictive Approaches for Safer Urban Environments, ran for three and a half years from November 2022 and closed at the end of July 2026. It was funded by Horizon Europe and UK Research and Innovation with a total budget of roughly €3.37 million, and delivered by eleven partners across three pilot cities: Athens, Valencia and the West Midlands.
The problem it set out to solve
Cities already have traffic simulation tools. They also have road safety assessment tools. What they have not had is a way to make the two talk to each other.
Existing simulation models are good at predicting congestion, journey time and vehicle throughput. They are considerably worse at predicting what happens to a pedestrian, a cyclist or an e-scooter rider when a junction is redesigned or a new mode arrives in the city. Safety assessment, meanwhile, has historically been retrospective: you count the crashes, then you fix the location.
PHOEBE’s premise was that a city planning a change should be able to model its safety consequences in advance, in the same way it models traffic flow.
The technical core of the project was the integration of iRAP’s ViDA road assessment platform with Aimsun Next traffic simulation, layered with behavioural modelling, mode-shift and induced-demand modelling, and telematics data. The consortium describes the output as a modular framework: cities can adopt individual components rather than the whole system.
What the three pilots looked at
The three use cases were deliberately different in scale and question.
Athens narrowed in on the city centre, where researchers from the National Technical University of Athens used the framework to assess the safety impact of the Great Walk pedestrianisation scheme. This is the closest the project came to evaluating a major, politically contested urban intervention.
The West Midlands took the opposite approach, applying the framework across a large geographical area rather than a single scheme.
Valencia examined the city’s main traffic arteries, and produced what is arguably the project’s most quotable result.
The Valencia finding
Researchers looking at Valencia found that introducing on-street parking led to an increase in illegal pedestrian crossings.
That single sentence is worth more than most of what gets published about predictive safety modelling, because it is exactly the kind of second-order effect that conventional assessment misses. A parking policy is not a pedestrian policy. It goes through a different department, gets justified on different grounds, and is evaluated against different metrics. Yet the parked vehicles change sightlines, lengthen crossing distances and remove the informal desire lines pedestrians were using, and the pedestrians respond by crossing where they should not.
No crash has to happen for a model to flag that. That is the entire argument for predictive assessment, demonstrated in one case.
Who built it
The consortium brought together eleven organisations: the European Institute of Road Assessment, the National Technical University of Athens, TU Delft, the Technical University of Munich, Aimsun, POLIS, FACTUAL, the Universitat Politècnica de València, OSeven, The Floow and iRAP.
That mix matters. It pairs academic modelling groups with the commercial owners of the actual tools (Aimsun, iRAP, telematics providers The Floow and OSeven) and with POLIS, the network that represents the city authorities who would have to use the result. Research projects that skip the last category tend to produce frameworks nobody adopts.
Findings were presented at the Road Safety and Simulation Conference in Naples in June 2026, with a closing webinar held on 30 June. PHOEBE also coordinated a joint paper on road safety data with its sister projects V4Safety and SOTERIA, drafted through the EU Road Safety Cluster and presented at the RTR Conference in Brussels in February 2026.
Why the EU funded this
The policy context is a target the EU is not currently on course to meet.
Around 19,000 people were killed on EU roads in 2020, and the bloc has committed to halving road deaths and serious injuries by 2030. Roughly 40 percent of road fatalities occur in urban areas, and the urban share is where the vulnerable road user problem concentrates.
Cities are also changing faster than assessment methods can keep up with. E-scooters arrived in most European cities within a two-year window and were regulated afterwards. Low-traffic neighbourhoods, bus lane conversions and pedestrianisation schemes are being implemented at pace. Each one redistributes risk, and under a retrospective model the redistribution only becomes visible once someone has been hurt.
What happens now
This is the open question, and it applies to every Horizon project that reaches its end date.
PHOEBE has produced a framework, a dashboard linked to iRAP’s ViDA platform, and a set of deliverables published openly on the project website. The consortium describes the framework as a blueprint other European cities can adopt.
Whether they do is a separate matter. EU-funded road safety projects have a long record of producing sound methodologies that never leave the PDF. The variables that determine adoption are unglamorous: whether a city has staff who can run the models, whether the data requirements are realistic for a mid-sized authority, and whether the tools remain maintained once the funding stops.
The integration with a commercially maintained platform gives PHOEBE a better chance than most. ViDA and Aimsun Next will continue to exist and be supported regardless of the project’s end date, which is not something that can be said of a bespoke tool built inside a research consortium.
The measure of whether PHOEBE worked will not be its final report. It will be whether a city transport department somewhere runs a predictive safety assessment on a scheme in 2028 and changes the design because of what it shows.
