ARES OS allows users to transform automated experiments and carbon nanotube synthesis reactors into autonomous ‘research robots’ capable of directing and conducting their own research using artificial intelligence and automation. These ‘robots’ support human research efforts by analyzing various methods of discovery/data collection, identifying the most efficient ones and then learning the quickest way to accomplish certain tasks.
Maruyama recently coauthored a review of autonomous research systems and their role in advancing future developments in materials science. The article, which is the September cover story in the journal “Matter,” explores the current and future state of autonomous research systems and is currently available online at https://www.cell.com/matter/issue?pii=S2590-2385(20)X0010-5.
In essence, AI helps design experiments, answer research questions, execute experiments and analyze the results, which then helps design a new round of experiments.
“This closed-loop, iterative process is faster and more effective than traditional research, and enables scientists to tackle harder, more complex problems,” said Maruyama.
Dr. Daylond Hooper, lead software developer, explains that, “ARES OS is the first software to robustly combine device control, analysis and AI-based planning while employing the latest software engineering principles and technology. In essence, the software incorporates various AI algorithms that help determine the most suitable strategy for a given experimental problem.
Before this tool, “Only well-funded, state-of-the-art labs typically had resources for achieving the kind of experimental throughput needed to realize technological progress in a short amount of time,” said Maruyama. “That’s why our team wanted to make autonomous experimentation more widely accessible … to harness the power of crowd-sourcing citizen science.”
Ultimately, ARES OS frees humans from designing, analyzing and running countless individual experiments, so they can instead focus on defining broad research goals for the software to pursue independently. In the end, this means that “humans and autonomous research robots can team together to solve research problems more effectively than either could alone,” Maruyama explained.
“Autonomous Research Systems do not replace humans, but rather they provide a platform for the partnership between the human researcher and the robotic system – a human-machine team,” he said. “The beauty is that that the ARES OS software lowers the barrier to entry for everyone to build their own research robot.”
In the past, researchers wrote similar code just to get one job done, a one-time experimental process, for instance, which sometimes led to poor code reuse and limited adaptability to new experiments.
“Today, when most people try to build a research robot, they typically build code from the ground up, Maruyama said, noting that ARES’ incorporation of established software engineering approaches means that this unique program can be readily applied to many kinds of research problems.
Maruyama noted that, “While the platform saves time up-front, users still have the option to inspect and customize the source code, since ARES is open-source software. We hope that a community will build around ARES OS where groups will contribute their code so that everyone can benefit.
Originally created to address AFRL needs, the ARES OS features a core or a universal module with plug-in’s that connect to experimental tools or scientific instruments such as electron microscopes. After this link occurs, the software can design experiments autonomously and perform experiments in a rapid, precise manner, and learn how to achieve the research goal using AI and autonomy.
In 2010, scientists and engineers from AFRL’s Materials and Manufacturing Directorate and Autonomous Capabilities Team 3, in collaboration with the Air Force Institute of Technology, pioneered the development of the first autonomous research system for materials research. Since then, AFRL has expanded the application of autonomous research from the original work in carbon nanotube synthesis to additive manufacturing and flow chemistry.
In recent years, the software driving ARES evolved into ARES OS, a comprehensive, general-purpose, autonomous research software platform, with an adaptability that encourages application in many fields.
James Deneault, a research engineer who works with 3D-printing technologies at AFRL, sees the software platform as a game changer.
“The parameter space in 3D printing is huge,” he said noting that researchers spend hours printing failed parts before finding the right settings for a given material. This process is particularly difficult with new experimental materials,” Deneault said.
“Human intuition can only handle so many parameters at a time… but this kind of high-dimensional and repetitive tuning is something that computers are really good at, which is why this software is brilliant in its approach.”
For the average, everyday user, this software teaches cheap 3D printers to print better without all the manual calibration, Deneault said.
Within AFRL, ARES OS continues to see new research applications. Numerous institutions in the U.S. and abroad are also evaluating the software including national laboratories, academic institutions and industry partners. Sandia National Laboratories, Ares Materials Inc., Wake Forest Baptist Health, are on this list along with Carnegie Mellon University, Ohio State University, University of Buffalo, University of Missouri, Rice University, University of Connecticut and the University of Glasgow.
“We’re very excited to see so much interest,” said Maruyama, emphasizing the fact that making ARES OS open-source is synergistic. “By making it easier for people to do autonomous research, this helps us to improve the software for everyone,” he said.
To download the ARES OS, visit https://forms.gle/4CiZkqQDDfVz5Jho8 and complete the short survey. The file includes the 3D-printer application along with instructions on how to link this the software with most commercially available machines in the $250 to $300 range.
For more information on AFRL’s efforts advancing autonomous research and design tools, please visit: https://afresearchlab.com/technology/successstories/polymer-and-responsive-materials-team/
About AFRL
The Air Force Research Laboratory is the primary scientific research and development center for the Department of the Air Force. AFRL plays an integral role in leading the discovery, development, and integration of affordable warfighting technologies for our air, space, and cyberspace force.
With a workforce of more than 11,500 across nine technology areas and 40 other operations across the globe, AFRL provides a diverse portfolio of science and technology ranging from fundamental to advanced research and technology development. For more information, visit www.afresearchlab.com.
About the Author