Wanted: artificial intelligence (AI) and machine autonomy algorithms for military command and control
ARLINGTON, Va. – U.S. military researchers are asking industry to develop reliable and trusted artificial intelligence (AI), machine learning, and systems autonomy technologies for future command and control, surveillance, and logistics systems.
Officials of the U.S. Defense Advanced Research Projects Agency (DARPA) in Arlington, Va., issued a broad agency announcement (HR001122S0039) last week for the Assured Neuro Symbolic Learning and Reasoning (ANSR) project.
The ANSR program seeks breakthrough innovations in new hybrid AI algorithms for military command and control that deeply integrate symbolic reasoning with data-driven learning to create robust, assured, and trustworthy systems.
Researchers say hybrid AI algorithms able to acquire and integrate symbolic knowledge, and to perform symbolic reasoning at scale, will deliver robust inference, generalize to new situations, and provide evidence for assurance and trust.
Machine autonomy should be to hasten the tempo of military operations, reduce cognitive demands on the warfighter, and increase standoff distances to help keep warfighters safe, researchers say.
The prevailing trend in industrial machine learning research is towards scaling up to giga- and tera-scale models with hundreds of billions of parameters to improve accuracies and performance.
These trends are not sustainable, however, because of the extremely high computational and data needs for training such models, and the inability of data- and compute-starved military applications to access cloud-scale computer resources.
U.S. military applications, in addition, are safety and mission-critical, and so need to operate in unseen environments, need to be auditable, and need to be trustable by human operators. Instead, the ANSR program seeks to develop new hybrid AI algorithms that are more conducive for use in real-world military applications.
ANSR seeks to modify training and inference procedures to interleave symbolic and neural representations that will result in representations that are grounded in domain-specific symbols, and can include prior knowledge and domain-specific rules.
The program will demonstrate assured execution of an unaided intelligence, surveillance, and reconnaissance (ISR) mission to develop a common operating picture of a dense urban environment. The autonomous system will carry an effects payload to reduce sensor- to-effects delivery time.
The capabilities required of the autonomous system today are not achievable by state-of-the-art machine learning or standalone symbolic reasoning systems. The ANSR program has four technical areas: algorithms and architecture; specification and assurance; platforms and capability demonstration; and assurance assays and evaluation.
The guiding challenge for the program will be the assured execution of an unaided ISR mission in a dynamic dense urban environment.
Algorithms and architecture seeks to develop and model new AI algorithms and architectures that deeply integrate symbolic reasoning with data-driven machine learning. Specification and assurance will develop an assurance framework and methods for deriving and integrating evidence of correctness and quantifying mission-specific risks.
Platforms and capability demonstration will develop use-cases for engineering applications of hybrid AI algorithms for robust and assured performance. Assurance assays and evaluation will develop an assurance test harness with adversarial AI, and evaluate these technologies in systems.
The program will seek to demonstrate an autonomous ISR mission that requires the warfighter only to identify the area of interest, and enable AI to analyze the needs for navigation and actions against the threats.
Companies interested should upload abstracts no later than 13 June 2022 and proposals no later than 26 July 2022 to the DARPA BAA Website at https://baa.darpa.mil.
Email questions or concerns to Sandeep Neema, the ANSR program manager, at [email protected]. More information is online at https://sam.gov/opp/0c28fb55fcb446dc95ed3337b385b36c/view.
John Keller | Editor-in-Chief
John Keller is the Editor-in-Chief, Military & Aerospace Electronics Magazine--provides extensive coverage and analysis of enabling electronics and optoelectronic technologies in military, space and commercial aviation applications. John has been a member of the Military & Aerospace Electronics staff since 1989 and chief editor since 1995.
