Learning optimal strategies for line-of-sight based games

Richard Tsai
University of Texas at Austin
Mathematics

We present a few non-cooperative games that involve the line-of-sight of the players. The games are set in non-simply connected domains, i.e., domains that have obstacles blocking the lines-of-sights. We present Monte-Carlo-Tree Search (MCTS) based algorithms for optimally playing the games. The search algorithms are aided by deep learning and local numerical computation of optimal strategies for the players.


Back to Workshop I: High Dimensional Hamilton-Jacobi Methods in Control and Differential Games