Adversarial Autonomy

The term “Adversarial Autonomy” refers to an autonomy engine that must engage with and against an intelligent human operator/s, who is trying to actively defeat the group of vehicles, for continuous extended periods of time in a relatively unconstrained conditions. There are a number of existing systems that do engage with/against humans; however, these engagements are relatively short in nature and fairly structured in their operations. This is a significantly different planning structure.

Game Theory has been applied to many situations. One of the most recognized examples is the computer “Deep Blue” that played, and beat, Garry Kasparov in chess.   Prior to the Deep Blue – Kasparov match, developers had spent decades trying to, unsuccessfully, build a computer program that could beat humans in chess.       Many different techniques were attempted including expert systems, but none were successful. Computers were only able to finally beat humans when an efficient Game Theoretic Technique was adopted, which allowed computers to play to their strength, computational power.

Game theory is “the study of mathematical models of conflict and cooperation between intelligent rational decision-makers.” (Myerson, 1991) This mathematical theory has been used for various applications ranging from solving board games to modeling competitive behaviors in economics to political war bargaining. For this paper, we will be specifically addressing Combinatorial Game Theory (CGT) used to solve a zero-sum (i.e. “I gain what you lose”) strategic game between several players: the red team, bystanders, other team members.