Army shines light on AI, machine learning research at virtual conference

By U.S. Army CCDC Army Research Laboratory Public AffairsMay 12, 2020

Army shines light on AI, machine learning research at virtual conference.
Army shines light on AI, machine learning research at virtual conference. (Photo Credit: U.S. Army) VIEW ORIGINAL

ADELPHI, Md. -- Army researchers organized and chaired a scientific conference to advance artificial intelligence and machine learning -- virtually -- to help build the foundation for the future force.

The 2020 International Society for Optics and Photonics Defense and Commercial Sensing Conference on Artificial Intelligence and Machine Learning for Multi-Domain Operations Applications II, slated to take place in Anaheim, California, April 26-30; however, given the current global environment, the event quickly transformed into a free virtual conference online.

“The goals of this conference are to promote understanding of near-term and far-term implications of artificial intelligence and machine learning for multi-domain operations, and to gain awareness of research and development activities,” said Latasha Solomon from the U.S. Army Combat Capabilities Development Command’s Army Research Laboratory. “The laboratory understands the operational importance of AI for the Army and was inspired to organize the first conference in the spring of 2018.”

Solomon and Dr. Tien Pham, also from the laboratory, led the effort, with over 90 invited and contributing papers as part of the SPIE Defense + Commercial Digital Forum, April 27 to May 1.

AI and machine learning are building blocks for the Army Modernization Priorities. Officials plan to leverage future AI and machine learning capabilities to provide intelligence support. As part of the Army’s number one modernization priority, an AI Hub will lead development into an intelligence support platform designed to identify, assess and prioritize threats based on real time updates from all available sensors, both manned and unmanned.

The technology also shows promise for automated threat recognition, which may help to develop systems to support an automated vehicle and fleet of vehicles, which would support the Next Generation Combat Vehicle priority.

For the virtual conference, the U.S. Army Futures Command’s Futures and Concept Center and the Defense Advanced Research Projects Agency’s Tactical Technology Office were among the 71 presentations and papers.

Participants included more than 3,600 registered DCS conference attendees and contributors spanning multiple diverse organizations to include academia, industry, the Army, Navy and Air Force, coalition partners and government agencies, each possessing a unique skillset to address the challenging path forward of artificial intelligence and machine learning, or AI/ML, for multi-domain operations, commonly referred to as MDO.

MDO describes how the U.S. Army, as part of the joint force -- the Army, Navy, Air Force and Marines -- will solve the problem of layered standoff in all domains -- land, sea, air, space, and cyberspace.

In addition to the laboratory, participating organizations included the University of Illinois, Air Force Office of Scientific Research, IBM Thomas J. Watson Research Center, IBM United Kingdom Limited, Defence Science and Technology Lab, Cardiff University, Mitre and Discovery Lab-Global.

The lab invited Dr. Katie Rainey from the Naval Information Warfare Center in San Diego, California, to be a co-chair this year. Rainey chairs the annual meeting on Naval Applications of Machine Learning and brings participants from the Navy and West Coast to participate in this conference.

Authors uploaded pre-recorded briefs and responded to posed technical questions and comments throughout the duration of the conference. This format worked exceptionally well, Pham and Solomon said, as it gave many attendees who may not have otherwise had the opportunity to see the brief, due to parallel sessions, an opportunity to view presentations and engage one-on-one with authors.

This was year two of the laboratory-led conference, inspired in response to the lab’s 2016 Artificial Intelligence and Machine Learning Essential Research Area, which explored AI that is unique to military operations at the tactical edge and are not addressed by the commercial sector.

This collaborative forum provided a unique opportunity for attendees and subject matter experts to realize and discuss relevant capability gaps and emerging technical advances in an unclassified setting while often resulting in opportunities to develop future collaborative efforts, Solomon said.

The laboratory session chairs and authors used the venue as a platform to influence where academic and businesses invest their research, focusing science on Department of Defense priorities, she said.

AI is one of nine Army Research Priority Areas, and is foundational to all six of the U.S. Army Modernization Priorities.

“To solve some of MDO’s most critical problems, the future force requires the ability to converge capabilities from across multiple domains: land, air, sea, space, cyber, spectrum, information, etc., at speeds and scales beyond human cognitive abilities,” Pham said. “Specifically, at the tactical edge, future military operations will involve teams of highly-dispersed warfighters and agents, both robotic and software, operating in distributed, dynamic, complex, cluttered environments with near-peer/peer adversaries.”

According to the organizers, military applications are frequently distinct from commercial applications because of rapidly changing situations, limited access to real data to train AI/ML, noisy, incomplete, uncertain and erroneous data inputs during operations and peer adversaries that employ deceptive techniques to defeat algorithms.

“Most current research in AI/ML is accomplished with extremely large collections of relatively clean, well-curated training/operational data with little background noise and no deception,” Pham said.

In response, over the past two years, the lab has organized conference sessions to address the following research topics:

  • Learning and reasoning with small data samples, dirty data, high clutter and deception 
  • Autonomous maneuver in complex environments 
  • Federated/distributed AI/ML 
  • Human agent teaming 
  • AI-enabled situational awareness and context-aware decision making 
  • Resource-constrained AI processing at the point-of-need 
  • Adversarial machine learning 
  • Explainable and AI-enabled Analytics 
  • Novel AI/ML algorithms, frameworks and applications 

Two highlights from the conference include the most viewed and downloaded presentations.

The first is the plenary brief, “Artificial intelligence for Maneuver and Mobility Essential Research Project,” which details the lab’s approach to addressing the future contested operational environment. Authors include Stuart H. Young, DARPA Tactical Technology Office and ARL, and John Fossaceca, CCDC-ARL.

As detailed in the brief, the future operational environment will be contested in all domains in an increasingly lethal and expanded battlefield, conducted in complex environments against challenged deterrence.

According to the authors, in order to prevail in the MDO phases of dis-integration, exploitation and re-entry to competition, the Army will need to employ teams of highly-dispersed warfighters and agents, both robotic and software, to include Robotic Combat Vehicles. To enable this human-agent teamwork, the Artificial Intelligence for Maneuver and Mobility Essential Research Project aims to revolutionize AI-enabled systems for autonomous maneuver that can rapidly learn, adapt, reason and act in MDO.

The authors said that this capability will enable autonomous vehicles to team with Soldiers more seamlessly, reducing Soldier cognitive burden; conduct reconnaissance to develop the enemy situation at standoff, creating options for the commander; and enabling the next generation of combat vehicles to fight and win against a near-peer adversary.

The other most viewed and downloaded brief, The Multi-Domain Effects Loop: From Future Concepts to Research Challenges,” discusses research challenges and enablers that underlie the design of effective MDO in a future defense theater marked by a large operation scale, rapid dynamics, heterogeneity, distribution and a contested/adversarial environment. Authors include Adam Taliaferro, FCC, Paul Sullivan, CCDC-ARL, Tarek Abdelzaher, University of Illinois, and Stephen Russell, CCDC-ARL.

In this brief, the authors presented future concepts that embody such challenges as the presence of near-peer adversaries and the need for tight coordination among large numbers of heterogeneous assets from multiple domains. As a running case study, research solutions to the above challenges are described from the lab’s recently launched Collaborative Research Alliance on the Internet of Battlefield Things.

Despite the challenges that arose, Army researchers and their partners said they successfully met to share and explore research that will ultimately benefit future Soldiers, who will need innovative technology to win the fight anywhere, any time.

For additional information, SPIE published a website on the conference.

CCDC Army Research Laboratory is an element of the U.S. Army Combat Capabilities Development Command. As the Army's corporate research laboratory, ARL discovers, innovates and transitions science and technology to ensure dominant strategic land power. Through collaboration across the command’s core technical competencies, CCDC leads in the discovery, development and delivery of the technology-based capabilities required to make Soldiers more lethal to win the nation’s wars and come home safely. CCDC is a major subordinate command of the U.S. Army Futures Command.