Efficient Algorithms for Computing Stackelberg Strategies in Security Games.

摘要

Game theory provides a framework for modeling a wide range of security and defense problems. This project focuses on Stackelberg strategies, which are optimal when one player can commit to a (possibly randomized) strategy before the other player moves. For example, a defensive unit can commit to a randomized patrolling pattern to deter attacks. This project explores new approaches for efficiently computing Stackelberg strategies in realistic security domains with exponentially large strategy spaces. Potential impacts of this research include increased ability to compute optimal strategies for security and defense scenarios.