News | May 4, 2018
Artificial intelligence accelerating materials discovery at Air Force Research Laboratory
Robotics and machine learning can accelerate the development process, leading to innovative, new materials at a faster pace for military and industry
DAYTON, Ohio – To create the next generation of technology, researchers at the Air Force Research Laboratory are turning to machine learning, artificial intelligence (AI) and autonomous systems to exponentially increase the speed of materials discovery and lower the cost of technology.
“We need to reinvent the process of doing research,” said Dr. Benji Maruyama, a research scientist at the Materials and Manufacturing Directorate.“Research is typically a slow process, with the time for materials conception and discovery to actual implementation on a platform for warfighters typically taking anywhere from 20 to 30 years in time. Today’s technological environment offers an opportunity for researchers to partner with research robots to advance understanding of multidimensional problems more rapidly.”
Maruyama is the coauthor of “Accelerating the discovery of materials for clean energy in the era of smart automation,” which was recently published in Nature Reviews, and discusses an integrated AI approach to materials discovery focused on the implementation of autonomous research systems in the laboratory environment. Maruyama is also the creator of the AFRL’s Autonomous Research System (ARES), which uses AI and machine learning to conduct autonomous experiments designed to optimize the synthesis of carbon nanotubes, which have tremendous potential for next-generation energy technology.
Today’s research problems are complex and highly dimensional, Maruyama said, with advanced materials impacting the full spectrum of daily life from energy generation, power, transportation, aerospace and more.
This requires the continuous development of new materials that are able to meet increasingly demanding performance requirements.
Platforms such as ARES that can integrate AI with robotics and machine learning provide an opportunity to accelerate the development process, leading to innovative, new materials at a faster pace and ultimately, decreased cost, for Air Force and commercial uses.
“Machines can complete work in multi-dimensional spaces more quickly than human researchers,” said Maruyama. “For example, every time ARES conducts a new experiment, it can use AI to analyze the results and design the next best experiment to perform in closed-loop automation, resulting in hundreds of experimental interactions per day. By removing the need for the human researcher to conduct tedious lab work to collect data, it creates the opportunity for more human insight and creativity, leading to faster implementation of material solutions.”
According to Maruyama, next-generation material development augmented by autonomous research and AI systems will likely experience “Moore’s Law” in which the rate of research will climb at an extremely high rate as the AI technology continues to improve, resulting in the discovery of innovative new materials faster than ever before.
Moore’s Law, a computing term developed in the 1970s, was founded on the premise that the overall processing power for computers would double every two years as technology improved. If the law holds true for research, says Maruyama, the material advancement potential is enormous.
“We started work on carbon fiber in the 1950s, and it took a long time for it to be used on aircraft and platforms. We still have a ways to go before we even realize carbon fiber’s full potential. If we can continue to update our understanding of the AI and autonomous systems environment, we will drive the speed of material innovation for the better,” Maruyama said.
Maruyama and his coauthors’ work has attracted widespread interest in both academia and industry, as the concept of uniting technological advances in automation, robotics, and computer science with research and development is a revolutionary perspective in the centuries-old field of scientific discovery. The ability to accelerate development and save cost by revolutionary approaches to long-term research is extremely enticing across the spectrum.
“We envision a new paradigm for materials discovery emerging over the next five to ten years or so,” said Maruyama. “We are changing the face of research.”
Over 130 inventions are already available from the Materials and Manufacturing Directorate. The technologies like the freshly patented moisture and temperature responsive polymers, and semi-transparent photovoltaic solar cells, are waiting to be commercialized by companies for military or non-military customers.
TechLink, the Department of Defense’s national partnership intermediary, helps businesses access new inventions from the Air Force through patent license agreements and cooperative research and development agreements.