A collaborative digital twin built on FAIR data and compute infrastructure
This work addresses the problem of inefficient collaboration and data management in scientific optimization for researchers and students, though it is incremental as it builds on existing FAIR and SDL concepts with a new implementation.
The paper tackles the challenge of accelerating discovery in self-driving laboratories by implementing a distributed framework that integrates FAIR data infrastructure with machine learning for collaborative optimization, resulting in a system where geographically dispersed researchers can share data and automatically update models to guide experiments. It demonstrates this with a frugal twin application for optimizing food dye combinations to achieve target colors, enabling accessible and inexpensive experimentation.
The integration of machine learning with automated experimentation in self-driving laboratories (SDL) offers a powerful approach to accelerate discovery and optimization tasks in science and engineering applications. When supported by findable, accessible, interoperable, and reusable (FAIR) data infrastructure, SDLs with overlapping interests can collaborate more effectively. This work presents a distributed SDL implementation built on nanoHUB services for online simulation and FAIR data management. In this framework, geographically dispersed collaborators conducting independent optimization tasks contribute raw experimental data to a shared central database. These researchers can then benefit from analysis tools and machine learning models that automatically update as additional data become available. New data points are submitted through a simple web interface and automatically processed using a nanoHUB Sim2L, which extracts derived quantities and indexes all inputs and outputs in a FAIR data repository called ResultsDB. A separate nanoHUB workflow enables sequential optimization using active learning, where researchers define the optimization objective, and machine learning models are trained on-the-fly with all existing data, guiding the selection of future experiments. Inspired by the concept of ``frugal twin", the optimization task seeks to find the optimal recipe to combine food dyes to achieve the desired target color. With easily accessible and inexpensive materials, researchers and students can set up their own experiments, share data with collaborators, and explore the combination of FAIR data, predictive ML models, and sequential optimization. The tools introduced are generally applicable and can easily be extended to other optimization problems.