Artificial Intelligence Joins the Team for Smarter & Faster Experiments

Researchers have developed new artificial intelligence methods to enable “autonomous experimentation.”

Image courtesy of Brookhaven National Laboratory
Compared to conventional manual grid measurement (left), the autonomous measurement method (right) produces a more accurate reconstruction of the sample’s shape (top row) by intelligently positioning measurements near features of interest (bottom row).

The Science

Scientists have developed a new artificial intelligence method that automates experiments. It autonomously defines and conducts the next step of an experiment without input from human researchers. The method works by creating a model that fits the data from an experiment. It then uses that model as the starting point for a process of continuously refining the model to fit with new data.

The Impact

Machine-guided studies can free scientists from managing the details of experiments. This allows scientists to focus instead on interpreting the results. Researchers demonstrated the new automated method on X-ray scattering experiments. The new method improved accuracy and produced results many times faster than conventional methods.


Scientific instruments are increasingly powerful, able to quickly acquire data and produce huge data sets. To make full use of these tools, researchers need ways to speed experiments and data analysis. The new SMART (Surrogate Model Autonomous expeRimenT) method is an autonomous decision-making algorithm that enables scientific instruments to autonomously explore scientific problems without human intervention. Using a statistical technique called Kriging, the SMART method selects what measurement to perform by calculating uncertainty, and thus the gain in knowledge expected with future measurements. In other words, it estimates what follow-up experiment will generate the most information about a scientific problem, then automatically launches that experiment. With each new experiment and each new measurement, the SMART method focuses on ever-smaller areas of remaining uncertainty with a scientific question.

After developing the SMART method, the researchers deployed the method for X-ray scattering beamline experiments at the Complex Materials Scattering operated in partnership by two Department of Energy user facilities, the National Synchrotron Light Source II and the Center for Functional Nanomaterials. The research team also included scientists from the Center for Advanced Mathematics for Energy Research Applications (CAMERA). The tests found that the SMART method was quickly able to focus on the area of interest in an image instead of the background, leading to improved accuracy and results up to six times faster than conventional methods.


Kevin Yager
Center for Functional Nanomaterials at Brookhaven National Laboratory

Marcus Noack
Center for Advanced Mathematics for Energy Research Applications (CAMERA) at Lawrence Berkeley National Laboratory


The work was partially funded through the Center for Advanced Mathematics for Energy Research Applications (CAMERA), which is jointly funded by the Advanced Scientific Computing Research (ASCR) and Basic Energy Sciences (BES) programs in the Department of Energy Office of Science. The research also used resources of the Center for Functional Nanomaterials and the National Synchrotron Light Source II, which are Department of Energy Office of Science user facilities.


Noack, M., et. al, “A Kriging-Based Approach to Autonomous Experimentation with Applications to X-Ray Scattering,” Scientific Reports 9  (2019). [DOI: 10.1038/s41598-019-48114-3]

Related Links

Smarter Experiments for Faster Materials Discovery: Brookhaven National Laboratory Newsroom Story

Highlight Categories

Program: ASCR , BES , SUF

Performer: SC User Facilities , BES User Facilities , CFN , NSLS-II