A new NATO STO research team aims to investigate how reinforcement learning could be used to support decision making within NATO, NATO Nations, and partners. The team, formed under the System Analysis and Studies (SAS) Panel, held its first in-person meeting this month at the STO Collaboration Support Office in Neuilly-sur-Seine, France.
Reinforcement learning (RL), also known as Approximate Dynamic Programming (ADP) in operations research, uses artificial intelligence (AI) to optimize decision making. Within the military context, most RL/ADP research has focused on autonomous agents, such as those used in unmanned aerial vehicles, communication networks and training networks. Recent reviews, however, suggest that RL/ADP has the potential to provide a significant advantage to decision makers across a broad spectrum of defence and security domains, including high-level command and control, military wargames, and force development.
The SAS research team (SAS-181) aims to investigate ways in which RL/ADP could be used to support effective and reliable decision making across the Alliance. At this month’s inaugural meeting, team members discussed terminology and definitions, and discussed gaps and trends in existing RL/ADP research. Over the next two years, the team will work to better understand the defence and security situations in which RL/ADP could be used to support decision making, develop best practices and recommendations, and foster a community of RL/ADP experts. The meeting brought together members representing 11 Nations and partners: Canada, Finland, France, Germany, Netherlands, Norway, Türkiye, United Kingdom, United States and Sweden.
Effectively applying RL/ADP requires cooperation and collaboration, covering a wide range of knowledge and expertise (both practical and theoretical). This is reflected in the NATO AI Strategy, which underscores NATO’s commitment “to collaboration and cooperation among Allies on any matters relating to AI for transatlantic defence and security”.
Click here to learn more about the work of the SAS Panel.
Published by SAS |
