The Road to Human-Level AI in Wargames: Algorithms for Implementing the Five Canonical Offensive Maneuvers in a CGF Environment and Making Tactical Decisions Drawing Upon Knowledge Acquired from a Dataset of Historical Battles

David Ezra Sidran

PhD Candidate
Department of Computer Science
University of Iowa

Abstract

After briefly describing human-level artificial intelligence we concentrate on the area of wargames, also known as a “computer generated forces” (CGF) environment. We present our research into the algorithms and underlying machinery necessary to implement the five canonical offensive maneuvers described in U.S. Army Field Manual FM 3-21, Section II, Forms of Maneuver (envelopment, turning movement, infiltration, penetrations, and frontal attacked) within a CGF environment. We also include descriptions of algorithms for calculating groups and flanks for sets of forces which are a necessary precursor to calculating the five offensive maneuvers. We will also demonstrate our TIGER program which is a fully functional GUI test bed for our research. Lastly, we will outline our current research in using a form of unsupervised machine learning known as conceptual clustering which will be used to match a given tactical situation to a tactical situation in a database of historical tactical situations that have been previously analyzed by subject matter experts (SMEs). Using this knowledge TIGER will then be able o recommend a optional course of action that conforms to SME doctrine.