AFRL posts embedded deep learning BAA
On April 27, the U.S. Air Force Research Laboratory (AFRL) posted an updated version of its Robust and Efficient Computing Architectures, Algorithms and Applications for Embedded Deep Learning broad agency announcement (BAA). For best consideration for FY 21 funding, AFRL recommends that white papers by submitted by May 29, 2020.
The Air Force Research Laboratory is soliciting white papers under this Broad Agency Announcement (BAA) for research, development, integration, test, evaluation and delivery of technology for computational capabilities with greater sophistication, autonomy, intelligence, and assurance for addressing dynamic mission requirements imposed by Command, Control, Communications, Computers, Intelligence, Surveillance and Reconnaissance (C4ISR) applications and Size Weight and Power (SWAP) constrained Air Force platforms. Of particular interest are advanced efficient computing architectures and systems, as well as robust algorithms and applications. The overarching objective is to achieve orders of magnitude improvement in size, weight and power for deploying robust artificial intelligence and machine learning (AI/ML) capabilities in an embedded computing environment. More detailed objectives are as follows:
Unconventional computing architectures are necessary to achieve advanced and new capabilities in pattern recognition, event reasoning, robust decision making, adaptive learning, and autonomous tasking for energy efficient agile Air Force platforms. A major focus area is neuromorphic computing, i.e., brain-inspired computing architectures, models, algorithms and applications. Neuromorphic computing research encompasses multidisciplinary development of novel processors beyond the Von Neumann architecture. Unconventional circuits and architectures can contribute to advancements in this area. There is a need for in-hardware robust and efficient adaptive learning architectures and algorithms, as well as novel AI/ML algorithms optimized for neuromorphic computing hardware architectures.
AFRL’s intent is to develop and demonstrate innovative modular computing system architectures and applications to meet the Air Force’s need for future real-time embedded plug-and-play capabilities. Technologies and applications may include, but are not limited to, artificial intelligence and machine learning models and algorithms for big data analytics for multi-source and multi-modal sensor data, data fusion algorithms for situation understanding and sense-making, and autonomous decision making techniques. Modular designs should support interchangeable sensors and other devices, with automatic software reconfiguration based on the available resources. Data bandwidth requirements of future systems can be expected to significantly increase. Compute and interface methods should be selected that will be scalable accordingly. Optimizations for size weight and power (SWaP) will be a priority.
Full information is available here.