Go directly to the content of the page Go to main navigation Go to research

The Hauts-de-Seine department is trialling innovative software to forecast road traffic on the ring road around La Défense, Europe’s leading business district. The solution is then set to be rolled out elsewhere.

Every day, around 30,000 drivers use a 4 km artery located west of Paris – officially named the RD993, but commonly known as the La Défense ring road – where traffic jams are a common occurrence. In 2019, the local council for the Hauts-de-Seine department decided to innovate here in order to boost the appeal of Europe’s leading business district.

A call for projects was launched in partnership with the Paris La Défense local authority and the Centre for Studies and Expertise on Risks, Environment, Mobility, and Urban and Country Planning (CEREMA). The aim was to design, implement and prove the real-time success of innovative solutions to make traffic smoother and safer – not only around La Défense but also through duplication at any other mobility infrastructure.

The departmental council selected four applications, including the AGIT system for smart traffic analysis and management designed by Citeos Solutions Digitales, a VINCI Energies business unit that focuses on smart parking solutions.

The solution designed by Citeos produces highly precise traffic forecasts 15 minutes ahead of time using machine learning.

Currently at the trial phase, the model developed by Citeos and its partner Qucit produces highly precise traffic forecasts 15 minutes ahead of time. To do so, the solution harnesses machine learning technology connected to the traffic management centre for the interdepartmental public authority (EPI) for the Yvelines and Hauts-de-Seine departments, which uses regional traffic data (speed, flow, journey times, etc.) as well as various external data (schedules, bank holidays, school holidays, the weather and road network data).

Hybrid forecast models

The first attempts to formulate a mathematical theory for road traffic date back to the 1920s. But even now, a century later, no one scientific theory has managed to truly understand the real factors that influence road traffic. Current forecast models therefore harness a combination of empirical and theoretical techniques.

We know that traffic behaves in a complex and non-linear manner. Due to how individual drivers react, vehicles do not simply interact in accordance with mechanical laws; rather, they tend to form clusters which, depending on their density, produce shock waves of varying sizes.

Project manager at Citeos Solutions Digitales, Alexane Gondel, said, “standard forecasting tools, like statistical analysis based on physically measuring vehicle flows, are still not suitable for precisely analysing dynamic traffic in a dense and complex environment like the Paris La Défense ring road. As far as we know, there is currently no model fully based on machine learning that can be used in real time.”

The decision to carry out trials at La Défense is also beneficial in that the solution designed by Citeos can incorporate rapid changes to traffic flows, as well as more long-term trends created by drivers’ evolving habits. For example, the recent lockdowns and subsequent lifting periods in particular have led to drastic changes to existing behaviour.

An open and easily replicable solution

In addition to analysing traffic and forecasting congestion incidents, the software solution rolled out at La Défense could, over time, offer a wider range of services such as alarm configuration for operators, decision-making assistance, scenarios for adjusted traffic signal programmes and speed limits, and communication with drivers using variable message signs. Citeos provides clients with software that summarises analyses via an online interface where they can visualise the results of the trials.

The aim is to roll out the demonstrator project currently underway at La Défense across the entire department, supporting the regional low carbon mobility strategy. AGIT was therefore designed with easy replication in mind.

“The system can be reproduced in other places and environments for the assessment and regulation of traffic at different types of infrastructure, such as environmentally friendly transport routes. It’s also an open solution that can be applied to different forms of mobility. AGIT can therefore provide forecasts for MaaS or cycling route planners, so they can adapt recommendations for users in real time,” pointed out Alexane Gondel.