Self-Driving Lab (SDL)./Courtesy of Argonne National Laboratory

Artificial intelligence (AI) and robots are rapidly changing the landscape of scientific research with the "Self-Driving Lab (SDL)".

Researchers at North Carolina State University announced on 14th that they have developed a self-driving lab technology that can collect more than 10 times the data at ultra-fast speeds compared to existing methods. The research results were published the same day in the international journal "Nature Chemical Engineering."

The self-driving lab is an automated experimental system that combines AI and robotic technology with chemistry and materials science. When researchers present a goal, AI devises the experimental plan, and robots conduct the experiments and subsequently analyze the results to determine the next experiment. Research continues to progress without human intervention.

Until now, the self-driving lab has relied on an experimental method known as "steady-state flow." This involves mixing experimental substances and waiting until the reaction is complete before analyzing the results. This process can take tens of minutes to up to an hour for a single experiment. The system effectively comes to a halt until the characteristics of the results are analyzed.

The research team has eliminated inefficiencies by introducing a new "dynamic flow." This method allows experimental substances to flow constantly while collecting data in real-time. There is no need to wait for the reaction to finish, as data is saved every 0.5 seconds. If the reaction time is 10 seconds, 20 data points can be obtained.

Milad Abolhasani, a professor at North Carolina State University, explained, "Previously, experiments provided only a single frame of a picture, but now they can be viewed in real-time like a video. Thanks to this, the system keeps operating and continues learning."

With data being quickly obtained, the AI's predictive capabilities become more precise. It can more accurately determine which substances are promising, reduce failures, and reach goals efficiently. In fact, the research team noted that this system collected 10 times more data in the same timeframe compared to existing methods.

The AI system identified optimal candidate substances right after training. With fewer experiments, chemical substances used are also reduced, leading to less waste. Professor Abolhasani stated, "This technology not only enhances research efficiency but also has positive environmental impacts," adding, "While the speed of experiments is important, how sustainably we conduct research is also key to the future of science."

AI-operated laboratories are reducing expenses and accelerating the pace of commercialization. On the 3rd, undergraduates from the University of Toronto unveiled a self-driving lab that can be built for under $500 (approximately 690,000 won). They created an initial version with LEGO blocks and improved the structure over three years. Information on required components, software, and assembly methods has been made freely available online so that anyone can replicate it.

Jason Hattrick-Simpers, the professor who supervised the research, noted, "The gap between those who have the opportunity to participate in science and those who do not may widen," stressing the need for low-cost self-driving labs that are accessible to everyone.

References

Nature Chemical Engineering (2025), DOI: https://doi.org/10.1038/s44286-025-00249-z