The integration of artificial intelligence (AI) and big data is paving the way for improved accident prevention and infrastructure management. This is the focus of the Road4Safe project, coordinated by the GeoTech group at the University of Vigo and involving 18 partners from 12 countries. The initiative aims to enhance road safety by predicting and preventing accidents before they occur, utilising AI and big data.
To achieve this, the project will develop a "digital twin" of road traffic and infrastructure, which will integrate data from multiple sources, including connected vehicles, infrastructure sensors, environmental data, and historical records. This will enable real-time risk predictions displayed on local and network-wide hazard maps.
Two complementary pilot projects will be implemented: one in Austria, focusing on critical scenarios in tunnels and entry and exit ramps, and another in the Bizkaia Corridor in Spain, which will assess risk prediction on a network scale. Upon completion, the technologies developed will provide a comprehensive solution for proactive safety management, aligning with the EU's "Vision Zero" goal.
Pedro Arias, the project's principal investigator and a professor at the University of Vigo, emphasised that the project's core focus is prevention, aiming to anticipate accidents. He noted that secondary roads and transitional sections, such as motorway access and exits or tunnels, are prioritised due to their higher accident severity caused by speed differentials and lower design standards. In 2023, approximately 25,600 people died on EU roads, with 56% of these fatalities occurring on secondary roads.
Main areas of focus
According to Arias, the project will address several key areas. Firstly, real-time traffic management will be developed to anticipate congestion, detect erratic driving, provide collision warnings, and recommend safe speeds for specific road sections. Another focus will be on infrastructure, identifying sections where risks are due to the road itself and proposing long-term measures. The project will also simulate how infrastructure deterioration and various maintenance strategies affect safety.
Connected mobility will also be a key area, particularly in terms of vehicle-infrastructure communication to relay warnings to drivers, along with designing interfaces that make these messages clear and comprehensible.
Project details
The University of Vigo coordinates a consortium of 18 partners from 12 countries, including five associated entities from Australia, the United States, and Singapore. Among the 13 beneficiaries are universities such as La Coruña, Chemnitz, Delft, and Queensland, as well as companies specialising in cutting-edge transport technologies, including the British firm HAAS Alert and the Hungarian company Commsignia. Spanish infrastructure companies Cemosa and Xouba Ingeniería are also involved, along with other stakeholders like Austria's public highway corporation Asfinag, the International Road Federation (IRF), and the Tecnalia Foundation. This broad collaboration ensures comprehensive coverage from research to real-world application.
The project officially commenced last week with a meeting in Vigo and is funded by the European Commission's Horizon Europe programme. It has a duration of 42 months and a budget exceeding €5 million, with approximately €862,000 allocated to the University of Vigo.
Digital twin and real-world pilots
Arias explained that current road safety management is largely reactive, addressing issues only after accidents occur. Road4Safe seeks to reverse this approach by anticipating risk situations to prevent accidents. The project will create a "digital twin" of the road and traffic, a virtual replica that continuously receives real-time data from various sources, including connected vehicles, sensors, cameras, and LiDAR installed along the roads, as well as weather information, road inventories, satellite images, and historical accident data.
AI will play a crucial role in several aspects, including integrating and refining these fragmented data sources, predicting risks at two scales—network-wide risk maps for medium to long-term planning and real-time predictions for specific segments (for example, anticipating traffic congestion in a tunnel or detecting erratic driving). AI will also assist in decision-making through a support system that proposes and prioritises measures for road managers, such as recommended speeds, access regulations, detours, or infrastructure improvements, simulating their effects in the digital twin before implementation.
Arias stressed the importance of ensuring that AI is reliable, with explainable decisions, free from biases that could disadvantage certain groups (such as pedestrians or cyclists), and compliant with European data protection and AI regulations. All of this will be validated through two real-world pilots: the Rustenfeld tunnel on a motorway south of Vienna, which handles around 85,000 vehicles daily, and the Connected Corridor of Bizkaia, a network of over 1,200 km of roads.
Contributions from the applied geotechnologies group
In addition to coordinating the overall project, the Applied Geotechnologies group at the University of Vigo will lead the work area focused on data collection, harmonisation, and sharing, which forms the foundation for all models. Specifically, Arias explained that they will concentrate on three aspects: identifying and integrating infrastructure and contextual data (road inventories, cameras, satellite images, weather sensors, historical accident data) and developing automated processes to clean, georeference, and upload this data to the project's common data platform. A second focus will be on developing AI tools to automatically extract road safety-relevant features from data collected by sensors, such as images, 3D point clouds obtained with LiDAR, or data from vehicles. Each result will include a reliability measure and a natural language explanation. The third objective is to create AI agents that autonomously select the most useful data and the best processing methods for predicting risk for each road segment, explaining their reasoning.
