Section: Computer Science and Engineering
Tutor: ALIPPI CESARE Major Research topic
:Adversarial team games
Advisor: GATTI NICOLAAbstract:
The analysis of adversarial team games is a largely unexplored topic in the algorithmic game theory literature. These game models are key to describe strategic interactions between rational agents that may organize themselves in a team (a group of agents sharing the same utility function).
We first analyze normal-form games in which a single team plays against a single adversary. The different possible correlation capabilities of the team lead to different solution concepts. In particular, when teammates cannot communicate during the execution of the game, the Team-maxmin equilibrium is the solution concept prescribing optimal strategies for the team. We study the theoretical properties of this solution concept, analyzing classical inefficiency indexes and introducing the Price of Uncorrelation as a way to characterize the value of intra-team communication. We devise algorithms to find and/or approximate the equilibrium, providing both a theoretical and experimental evaluation.
Then, we provide a study of extensive-form adversarial team games. Specifically, we introduce new correlation capabilities for teammates and study the corresponding solution concepts. Also in this setting, a theoretical and computational characterization of the different equilibria is provided.