Computer simulations are massively used in scientific research and in the industry for the analysis, design and optimization of physical systems. Over the last decades, Artificial Intelligence (AI) and Machine Learning (ML) methods have successfully entered science and engineering workflows to match the growing demands for fast and accurate physical models, thanks to a combination of improvements in the algorithms, computational power and data assimilation techniques. For example, AI and ML have contributed to advance weather prediction and the simulation of complex fluid flows. However, limitations of purely data-driven methods have emerged as concerns their generalization capabilities and their intelligibility.

To overcome these limitations, ALPS project (AI-based Learning for Physical Simulation), funded by ERC, proposes an original approach combining ML methods and mathematical modeling for the development of new algorithms that are able to automatically learn models of physical systems from experimental data. To efficiently handle the computational cost associated with the proposed methods, the algorithms will be implemented in a new software platform that seamlessly integrates automated model learning and high-performance simulation.
The methods developed in this project will be employed to address scientific challenges in human health, sustainable energy science and technology, and soft robotics. In particular, we envision new scientific discoveries in the problem of tumor growth, where accurate mathematical models are still elusive and could provide the basis for new treatment strategies. Further, we will use the algorithms to derive effective reduced-order models for model-based control in soft robotics and to tackle design, optimization and control problems in engineering.

FUNDER:  ERC - European Research Council 
GRANT NO.:  101039481
PERIOD AND TIME:  2022-2027 
FUNDING:  1,3 M€
COORDINATOR: Alessandro Lucantonio