Data Science
The decision-making capabilities of road Digital Twins are enabled by advances in data science and AI.
However, the use of AI within Digital Twins should go beyond black-box models that provide condition assessments or predictions without explaining the reasons behind them. The future of trustworthy AI for roads lies in developing new mathematical techniques, derived from areas such as probabilistic machine learning and physics-informed neural networks.
These methods need to be developed and evaluated using industrial data provided by our FUTUREROADS partners, ensuring that they address real-world challenges, including the complexity, uncertainty, and imperfections inherent in infrastructure data.
The Future Roads Programme Summary provides examples of research towards this aim. They include approaches for combining physics-based modelling and data-driven analysis for structural health assessment, developing probabilistic Digital Twin frameworks for understanding road conditions and supporting predictive maintenance, and enabling transparent and reliable decision-making throughout the lifecycle of road assets.
You can find details of the Future Roads Fellows research in Digital Twins as part of the Future Roads Programme Summary:
In the Digital Roads Prosperity Partnership, data science faces challenges related to data integration from diverse sources that are heterogeneous in fidelity, volume, coverage, quality, and levels of information available from material-based sensing systems, robotic systems, and digital twins. This research theme is also driven by the need to ensure sound quantification of uncertainty in data, providing the means to couple survey data with computational digital twins and robotic processes.
Data Science is using road survey data to provide road health diagnosis, through an automated decision support tool that translates the real-world data obtained from road surveys to their appropriate treatment plan, assisting Road Authorities in planning repair and maintenance work on the network.