Researchers strive to make AI a Soldiers’ trusted ally
ABERDEEN PROVING GROUND, Md. -- Movies and TV shows like to depict artificial intelligence as a terrifying, futuristic enemy whose unemotional and ruthless actions impart fear onto others, but Army researchers believe autonomous systems may one day become a Soldier’s most valuable ally.
The U.S. Army Combat Capabilities Development Command’s Army Research Laboratory designated several research programs as essential for future Soldier capabilities. Of these major flagship programs, the Human Autonomy Teaming, or HAT, Essential Research Program focuses on the relationship between humans and AI and how to establish human-robot teams that can survive and function in complex environments.
“We’re entering a society where autonomous technologies, or AI-based technologies, are becoming more and more prevalent,” said Dr. Amar Marathe, program manager for the HAT Essential Research Program. “We see it in the commercial market and we’re starting to see those types of technologies move toward military use. And as those technologies grow and develop over the next several years, we expect an influx of those technologies on the battlefield, not just for the United States but across the world.”
According to the researchers, AI and machine learning technologies enable the processing of large volumes of data very rapidly, which allows warfighters to make decisions faster and with greater confidence. This increase in processing capability will also accelerate the pace of actions on the battlefield and result in the emergence of even more complex situations to consider.
While autonomous systems boast impressive processing speed and decision-making capacity, they work best in environments that are similar to the one in which they are trained. In contrast, humans take longer to process large volumes of data, but they can more easily adapt to new environments and unforeseen changes to the situation.
LISTEN: What We Learned Today Podcast- Essential Research in Human Autonomy Teaming
Given these disparate competencies, Army researchers have begun to analyze how to effectively unite Soldiers with autonomous technologies to form human-autonomy teams that can outperform, out-maneuver and out-adapt their adversaries.
“The purpose for HAT is really thinking about how Soldiers and AI-based technologies will work together in the future,” Marathe said. “This essential research program looks at how we can take our Soldiers and couple them with AI-based technologies that can process large volumes of data, pull in information from lots of different sources and provide suggestions at an unprecedented pace. How do we put those two together to maximize the effect of the overall force that we can deploy on the battlefield?”
In light of this effort, the program seeks to understand the development of shared understanding of the situation between team members and how to confer this ability to intelligent technologies.
As Soldiers encounter any number of situations from potential enemy contact to difficult terrain, they can communicate with each other and use their shared experiences to overcome obstacles. Army researchers hope to impart that very same capability onto autonomous systems so that they can also communicate with their human partners and use the information given to them to not only recognize different scenarios but to also shape their behavior appropriately.
“In the end, AI-based technologies are still rule-based systems—they work on data that they have seen before and tend to fall apart when you change the environment in which they operate in,” Marathe said. “But, if we’re able to team both Soldiers and AI systems effectively, we can start to capitalize on the unique strengths and contributions of both and mitigate against some of the weaknesses that they may have individually.”
The Human Autonomy Teaming Essential Research Program consists of several project teams geared toward solving issues surrounding shared understanding, the coordination of action and the capability for rapid adaptation.
Army researchers have also formed partnerships with academic institutions, such as Carnegie Mellon University, Columbia University and the Massachusetts Institute of Technology, to develop novel theories of team-level processes for heterogeneous human-agent teams.
In addition, Marathe and his colleagues are working closely with engineers at the Combat Capabilities Development’s Command Ground Vehicles Systems Center and the Armaments Center to transition the laboratory’s research products into prototype army systems.
“In the short term, we’re focusing on solving fundamental challenges for specific Army applications to the Next Generation Combat Vehicle and Soldier Lethality at a squad or platoon level and identifying ways to improve the Soldiers’ situational awareness, their coordination and their capability to adapt in those specific application spaces,” Marathe said. “In the long term, we want to look at how the technologies that we are developing at these levels for ground vehicles and dismount Soldiers and determine how we can stitch everything together into a coordinated action across multiple domains.”
So far, the essential research program has made considerable success in the development of new tools to help Soldiers better understand the behavior, intentions and goals of autonomous systems. These tools demonstrated a dramatic improvement in mission effectiveness during simulated military operations and are currently being integrated into prototype vehicles.
The team also devised a special set of approaches to teach new behaviors to autonomous systems through human interaction. Instead of requiring a computer programmer to reprogram the autonomous system, the system can modify its behavior over time and learn to perform the desired behavior based on the feedback it receives from a person.
Other endeavors have focused on making autonomous systems easier for a human to understand through smaller, more subtle changes. For example, researchers have taken into consideration how the presentation of images can impact the user’s ability to comprehend its contents during the development of new aided target recognition systems and helmet mounted displays.
“Right now, we’re starting to look at creative ways in which we can pair groups of Soldiers with AI to brainstorm new options,” Marathe said. “One of the biggest challenges is the fact that the creative thinking space is largely a human-only problem—AI does not play in that domain effectively yet. But if we’re successful, I think we can bring the human autonomy teaming challenge from a low-level interaction to something much higher in mission command across much broader teams.”