AI helps human experts improve road maintenance

Traditional road maintenance planning is usually based on procuring specific measures at the lowest possible price. However, when contractors are awarded design, construct and maintain (DCM) contracts of up to 15 years, there is a greater need for more dynamic and sustainable methods. A new project is using interaction between AI and human expertise to improve road maintenance planning.

A new research project is studying the use of artificial intelligence (AI) to improve the planning and quality of road maintenance. Managed by Thomas Lundberg at VTI and Richard Nilsson at Skanska, the two-part project will explore the use of AI and machine learning to analyse large datasets that are more extensive than simply road surface measurements. The main purpose of the project is to determine which technology provides the best results and how it can be applied to long-term DCM contracts.

The project focuses specifically on Skanska’s DCM contract for the E22 highway through the counties of Kalmar and Östergötland. The contract combines traditional ocular inspections and advanced technology: mobile apps document damage on a weekly basis, deflectometers and ground-penetrating radar mounted to HGVs assess pavement strength and road structure, and VTI’s new advanced measurement vehicle – the VTI Mobile Research Platform – detects cracks and surface damage down to millimetre level. Evaluation is carried out using two different methods that are compared against one another, one traditional and the other data-driven.

Data is also retrieved from several open sources to create a more complete picture, including the National Road Database (NVDB), which contains data on the entire Swedish road network, and PMSv4, a database of road surface measurements of state-owned roads dating back to 1987. The project also uses soil moisture maps, soil-type maps and meteorological data on precipitation and temperature, as well data from sensors that measure frost boundaries on roads. All of this data is collected, positioned and analysed with the aid of AI models. The results are then reviewed by researchers at RISE Research Institutes of Sweden, who search for patterns and trends on which to base future decisions.

“One vital element of the project is exactly this combination of human expertise and automated analysis, the so-called human-in-the-loop (HITL) approach. While the goal is to automate parts of the process, it is crucial to balance the models against experience-based assessments,” Lundberg emphasises. The project also underlines the importance of being able to explain why certain damage occurs – and proposing measures rather than simply identifying that something is wrong.

The potential to change the industry

“We have high hopes for the project. The results can provide both contractors and road operators with better data on which to base decisions and lead to more durable roads at a lower cost. With a planning horizon of one year for a maintenance measure, by extrapolating conditions one year in advance hopefully the model can select the section of road with the highest priority while at the same time providing a versatile basis for choosing measures,” says Lundberg.

The objective is to develop a robust method for predicting the condition of the road. These forecasts can be used as a basis for selecting stretches of road for maintenance, the methods used and the ideal time for maintenance, from both durability and economic perspectives. The method is also expected to detect other types of damage and to improve response times. In the long term, the results may affect maintenance planning across the entire industry. A robust forecasting model should contribute to improving requirement specifications in future DCM contracts, increase awareness of how maintenance should be planned and lead to more cost-effective contracts – not just for the Swedish Transport Administration but also for municipalities and other road operators. Of course, ultimately the aim is to increase the working life of roads while reducing the need for urgent interventions.

The project is funded by the Swedish Competence Centre in Road Technology (KCV) and the Development Fund of the Swedish Construction Industry (SBUF). A final report is due in 2028.

Text: Christina Karlsson

Translation: CBG