Autonomous Experimentation: The Forefront of Scientific Research DX with AI and Robotics ~Analysis of Technology Trends from Grants, Papers, and Patents~

Astamuse Co., Ltd. (Headquarters: Chiyoda-ku, Tokyo; Representative Director and President: Ayumu Nagai) has comprehensively analyzed the technology domain related to autonomous experimentation, utilizing its proprietary innovation database (innovation and R&D information such as papers, patents, startups, and grants), and compiled a report on the trends.

What is Autonomous Experimentation?

Autonomous experimentation (autonomous/automated/automatic experimentation) is a research method that utilizes AI and robotic technology to conduct the entire R&D process, including experiment planning, execution, and results analysis, with reduced human intervention.

The origin of autonomous experimentation dates back to 2009, with the system "Adam" developed by Professor Ross King and his colleagues at Aberystwyth University (then) in the UK (Note 1).

Note 1: R. D. King et al., "The Automation of Science", Science 324, 85(2009).

https://www.science.org/doi/10.1126/science.1165620

"Adam" autonomously formulated 20 hypotheses about the relationship between yeast genes and enzymes and performed thousands of experiments to verify these hypotheses. In this verification, humans only performed material replenishment, waste disposal, and cleaning. Professor Ross King and his team conducted follow-up experiments on the results obtained by "Adam" and confirmed that the conclusions were consistent.

Scientific R&D traditionally relied on hypothesis generation based on researchers' tacit knowledge and experience, and painstaking manual trial-and-error (fine-tuning parameters, repeated synthesis, and evaluation). "Adam" was the first system in the world to complete the entire process from hypothesis formulation to experiment execution, results analysis, and hypothesis modification with minimal human intervention. "Adam" integrated AI and hardware to enable autonomous discovery, becoming the origin of the concept of "autonomous experimentation."

In 2019, with the advancement of machine learning algorithms, Professor Alán Aspuru-Guzik and his colleagues at the University of Toronto proposed the concept of "Self-driving Labs (SDL)" (Note 2).

Note 2: Alán Aspuru-Guzik et al.,” Next-Generation Experimentation with Self-Driving Laboratories”, Trends in Chemistry 1, 282(2019).

https://doi.org/10.1016/j.trechm.2019.02.007

The professors highlighted that a crucial element of autonomous experimentation is not merely performing pre-programmed experimental operations without human intervention (automation), but also evaluating experimental results, predicting the optimal conditions to try next, and modifying the plan autonomously without human intervention (autonomization).

With the rapid development of AI, autonomous experimentation is attracting attention as a field within "data-driven materials development," a new approach to materials development that uses information science for material exploration, prototyping, and evaluation.

Autonomous experimentation, through the automation and autonomization of R&D, is expected to bring value such as:

• Reduction of researchers' physical burden and alleviation of R&D talent shortages by replacing manual experimental operations with machines

• Shortening of R&D periods and reduction of waste by avoiding unnecessary experiments

• Inheritance of research know-how through the digitization of experimental operations

In this report, we focus on the automation and autonomization of hypothesis construction, experimental operations, and analysis in scientific R&D, utilizing AI and robots, and analyze the development trends of autonomous experimentation using Astamuse's unique database.

Research Budget Trends for Autonomous Experimentation Related Technologies

Grants (competitive research funds such as KAKENHI) represent investments in new approaches and research that have not yet led to published papers. Grant trends can be considered as indicators of R&D trends for technologies that require a relatively long time for social implementation.

From Astamuse's grant database, approximately 500 grants started since 2015, including keywords such as "autonomous lab," "laboratory automation," and "self-driving lab" in their research abstracts, were extracted.

Figure 1 shows the trend in the number of projects for the top 5 countries in terms of grant allocation for autonomous experimentation between 2015 and 2024. However, China is excluded due to non-disclosure of grant data, which likely does not reflect the actual situation. Also, recently published grant information may not yet be stored in the database, so recent figures may be underestimated.

Figure 2 shows the trend in research project allocation by country. Allocation amounts are evenly distributed over the project period and aggregated for each year. For example, a 3-year project with a budget of 30,000 USD is recorded as 10,000 USD for each year.

Over the 10-year period up to 2024, funding for R&D related to autonomous experimentation has been on an upward trend, which has intensified in the last five years due to the rapid growth of AI and information and communication technologies.

By country, the US is at the top in both the number of projects and research allocation. Particularly since 2020, it has significantly outpaced other countries in allocation. This is attributed to the early recognition of the concept of autonomous experimentation by the US government.

In the US, from July to October 2019, following the proposal of the concept of autonomous experimentation, meetings were held with American scientists and engineers to discuss opportunities for applying AI to scientific research. In 2020, the DOE (US Department of Energy) compiled the results into an "AI for Science" report (Note 3) for the government.

Note 3: DOE “AI for Science: Report on the Department of Energy (DOE) Town Halls on Artificial Intelligence (AI) for Science”

https://www.osti.gov/biblio/1604756

This report also mentions the concept of "self-driving lab," which is believed to have prompted the US administration to recognize the field of autonomous experimentation and the importance of using AI in research and experiments.

In 2025, the US announced the "Genesis Mission," an initiative to accelerate scientific discoveries using AI. Aiming to expand R&D productivity and impact within 10 years, the DOE announced an investment of approximately 320 million USD in projects that link the physical environment of laboratories with AI and robotics, such as laboratory automation and autonomous control of large-scale experiments (Note 4).

Note 4: DOE “Energy Department Advances Investments in AI for Science”

https://www.energy.gov/articles/energy-department-advances-investments-ai-science

The US is expected to further expand its investment in research projects to lead the field of autonomous experimentation.

The US's efforts have stimulated other countries, leading to the launch of large-scale funded projects in other nations as well. Below are examples of highly funded grants related to autonomous experimentation in various countries:

• MIP: BioPolymers, Automated Cellular Infrastructure, Flow, and Integrated Chemistry: Materials Innovation Platform (BioPACIFIC MIP)

  • Institution/Company: University of California-Santa Barbara

  • Grant Name/Country: NSF/USA

  • Research Period: 2020-2026

  • Allocation Amount: Approximately 24 million USD

  • Overview: Develop high-performance bio-derived plastics that surpass petroleum-derived materials by utilizing microorganisms such as yeast, fungi, and bacteria as "biological factories." Significantly accelerate material exploration, manufacturing, and evaluation through automated robotic synthesis and high-throughput experimentation.

• Autonomous Discovery of Advanced Materials

  • Research Institution/Company: UNIVERSITY OF SOUTHAMPTON and others

  • Grant Name/Country: CORDIS/EU

  • Research Period: 2020-2027

  • Allocation Amount: Approximately 11 million USD

  • Overview: Aims to build a platform for AI utilization in computational and experimental functional material exploration. In the computational process, AI and machine learning are used to automatically narrow down promising candidate materials from a vast chemical space. In the experimental process, AI-equipped mobile "robot chemists" autonomously conduct synthesis and evaluation experiments.

• AI for Chemistry: AIchemy

  • Institution/Company: University of Liverpool and others

  • Grant Name/Country: UKRI /UK

  • Research Period: 2024-2029

  • Allocation Amount: Approximately 7 million USD

  • Overview: The goal is to promote autonomous chemical experiments and reaction optimization using AI. It brings together experts in robotics and natural language processing. It aims to transform UK chemical research into an AI-driven model by developing data sharing infrastructure and fostering human resources.

• Creation of Multi-element Nanoalloys by Non-equilibrium Synthesis

  • Institution/Company: Kyoto University and others

  • Grant Name/Country: KAKEN/Japan

  • Research Period: 2020-2025

  • Allocation Amount: Approximately 630 million JPY

  • Overview: Research to create nanoalloys with atomic-level uniform mixing by momentarily creating a non-equilibrium state between normally immiscible metal elements under high temperature and pressure, then rapidly returning them to room temperature and pressure. Machine learning is used for material screening, and robot arms are utilized for automated synthesis equipment for material synthesis.

Analysis of Papers on Autonomous Experimentation Related Technologies

Papers represent research achievements by universities and academic institutions, reflecting technology trends with a shorter social implementation period than grants but longer than patents.

Approximately 2,900 papers published since 2015, including the same keywords in their titles or abstracts as for grants, were extracted. Figure 3 shows the annual trend of published research papers related to autonomous experimentation by the country of the first author's affiliation for the top 5 countries plus Japan, totaling 6 countries, published since 2015.

The US leads with more than double the number of papers compared to other countries. This suggests that the massive investments in R&D observed in the grant analysis are steadily yielding results.

Since 2022, China has taken the second position, pulling ahead of Japan, the UK, Germany, and Canada.

A "future estimation" analysis was performed on the paper population. Astamuse predicts nascent fields by calculating the annual trend of keyword occurrences to identify rapidly growing technological elements through a "future estimation" analysis. By tracing the evolution of keywords, it is possible to visualize technologies whose boom has passed and elemental technologies expected to gain prominence, as well as predict technology statuses such as nascent, embryonic, growth, and implementation.

Figure 4 shows the 10-year period from 2015 to 2024.