CICERO, N.Y. — The Air Force Research Laboratory (AFRL) Information Directorate has awarded SRC, Inc. a contract to develop neuromorphic on-chip autonomous learning (NOCAL) capabilities.
Working with AFRL and the College of Engineering and Computer Science at Syracuse University, SRC is developing advanced machine learning (ML) algorithms that leverage the extremely low-power processing and adaptive learning capabilities of neuromorphic hardware, the firm said in a news release. It did not disclose the dollar amount of the contract.
“SRC continues to innovate and pioneer work in the areas of ML, AI, and low-power computing,” Kevin Hair, president and CEO of SRC, said. “Our work with AFRL and Syracuse University on these novel algorithms will help our warfighters make informed decisions in the most challenging environments, allowing them to complete their mission successfully and come home safely.”
SRC is a nonprofit corporation headquartered in Cicero that focuses on areas that include defense, environment, and intelligence.
In traditional ML applications, systems require extensive training and rely on pre-evaluated models to accurately process data and provide actionable intelligence. SRC’s research under the NOCAL contract will concentrate on inventing, designing, developing, and analyzing technology to disrupt the current “state-of-practice.”
Artificial intelligence (AI) and ML capabilities on ultra-low-power neuromorphic hardware will “autonomously” adapt, learn, and support future missions without communications or offline pre-mission training.
These advanced AI and ML capabilities, coupled with neuromorphic hardware, can learn on-the-fly, evolve detection and classification models in real-time during missions, and are especially effective in environmentally and situationally constrained areas, SRC said.
SRC brings “decades of research and development expertise in machine intelligence and autonomy” to this new effort and has collaborated with AFRL in the past to develop the Agile Condor high-performance, embedded-computing architecture — a military system that applies ML algorithms and neuromorphic computing techniques to provide data analysis and processing, exploitation, and dissemination support for soldiers.