Development of NIMS-OS: general-purpose software enabling autonomous, automated experiments

1. A NIMS research team has developed the NIMS Orchestration System (NIMS-OS)—general-purpose software capable of closely coordinating a materials-search artificial intelligence (AI) system with an automated experiment system. NIMS-OS enables materials searches without human intervention and is now available to the public as open-source software.

2. Fundamental technologies are being developed with the goal of fully applying digital transformation (DX) to materials development. These include robotic experiment systems capable of automatically performing a series of tasks from materials synthesis to evaluation and materials- search AI systems designed to analyze existing materials data and suggest potentially promising materials worthy of further investigation. However, because these systems have been developed independently, it had proven difficult to coordinate them efficiently, undermining their overall performance.

3. This research team recently developed the NIMS-OS middleware, which is able to pair any materials-search AI and robotic experiment system to perform autonomous, automated materials- searches. This was achieved by enabling NIMS-OS to recognize these systems as separate modules. Three types of programs are pre-installed in NIMS-OS, including a Bayesian optimization algorithm. In addition, because the software is compatible with any materials-search AI system—even novel ones—it is readily able to run newly developed algorithms to operate the robotic experiment system.

4. The team then conducted a model experiment to verify the effectiveness of the NIMS-OS-coordinated system in identifying electrolytes suitable for use with lithium metal electrodes. In this experiment, NIMS-OS was arranged to control NAREE—NIMS automated robotic electrochemical experiments. The team confirmed close coordination between the AI and NAREE, enabling the successful execution of autonomous, automated experiments and a Bayesian optimization algorithm to identify potentially effective electrolytes.

5. The research team plans to add new abilities to NIMS-OS: the ability to coordinate with the many different types of robotic experiment systems and the ability to transfer data generated by a NIMS-OS-coordinated system to RDE, an automated data aggregation system developed by NIMS. Through these efforts to create data-driven materials development platforms, the team hopes to expedite the search for and discovery of innovative materials.

***

6. This project was carried out by a research team consisting of Ryo Tamura (Leader, Data-driven Algorithm Team, NIMS), Koji Tsuda (Invited Researcher, Data-driven Algorithm Team, NIMS; also Professor, Graduate School of Frontier Sciences, the University of Tokyo) and Shoichi Matsuda (Leader, Automated Electrochemical Experiments Team, NIMS).

7. This research was published in Science and Technology of Advanced Materials: Methods, an open access journal, on July 20, 2023, Japan Time. The NIMS-OS software is available to the public at the GitHub website (https://github.com/nimsos-dev/nimsos).