Multi-Agent Coordination for Strategic Maneuver with a Survey of Reinforcement Learning
Abstract
One promising avenue for implementing strategic maneuver to gain superiority over adversaries is through coordination of multi-agent systems (MAS) in future military operations. Recent work exploring the coordination of MAS has focused on the identification, classification, validation, implementation, and operationalization of emergent coordination through multi-agent reinforcement learning. Reinforcement Learning (RL) approaches can illuminate emergent behaviors through the exploration and exploitation of selected actions in a given environment, potentially leading to the inhibition of adversarial coordination which, in turn, can provide windows of opportunity across various intelligence, surveillance, target acquisition, and reconnaissance tasks. In this report, we present a brief overview of salient work in the RL domain and its potential applications for collaborative MAS for autonomous strategic maneuver.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 2021
- Accession Number
- AD1154872
Entities
People
- Anjon Basak
- Christopher D. Hsu
- Christopher Kroninger
- Derrik E. Asher
- Erin G. Zaroukian
- John A. Rogers
- John Fossaceca
- Luke Frerichs
- Michael R. Dorothy
- Piyush K. Sharma
- Rolando Fernandez
Organizations
- United States Army