Aligning AI and Robotic System for Autonomous Automated Materials Explorations

• Artificial Intelligence
• Automated Experimentation
• Center for Basic Research on Materials (CBRM)
• Data-Driven Approaches
• Machine Learning

A robotic experimentation system can be used to automatically synthesize and evaluate materials. Artificial intelligence (AI) can then analyze the evaluation results and propose next areas for investigation. Ryo Tamura developed the NIMS-OS middleware to seamlessly integrate AI and robotic experimentation systems, contributing to the advancement of data-driven materials development.

Using AI to make materials research 72 times faster

Data-driven techniques can significantly streamline and expedite efforts to identify new materials with desirable characteristics. Tamura is one of the researchers who have been developing AI tools for data analysis since the early days of the integration of data science into materials science. He has also been helping other materials researchers at NIMS and elsewhere by providing them with AI tools, substantially speeding up their searches for new materials.

In 2023, Ken Sakaushi (Research Center for Energy and Environmental Materials, NIMS) used AI tools developed by Tamura in his search for electrocatalytic materials capable of efficiently accelerating water electrolysis (NIMS NOW Vol.22, No.1, Leaders of the Future). Sakaushi succeeded in identifying a new, less expensive material as active as existing, platinum-based electrocatalysts in just a month. For reference, it has been estimated that the same comprehensive search process would have taken almost six years to carry out manually. This serves as a good example of how effective and powerful data-driven techniques can be when used in materials searches.

“AI-assisted materials searches require a variety of algorithms to achieve different objectives and identify different target materials,” said Tamura. “I’ve collaborated with a number of materials researchers specialized in different types of materials—including alloys, magnetic materials, solar cells and electrolytic materials—to develop AI tools tailored to their specific needs. Currently, more than 10 different algorithms are posted on our Data-driven Algorithm Team’s website, making them freely available for public use.”