Overview: |
In recent years, more attention has been paid to search-based AI solutions. Monte-Carlo Tree Search (MCTS) in particular has seen recent use for addressing more complex problems. But what do you do when your potential state space is so big, perhaps millions of possibilities big, to use traditional search-based algorithms? This session will describe the Hierarchical Portfolio Search AI system created by Lunarch Studios for their game, 'Prismata'. HPS is modular, robust, easily configurable, and can be built on top of a game's existing AI behavior library. David Churchill of Lunarch Studios will show how it can be adapted to fit almost any genre of game, including search-heavy ones like strategy games . He will show how HPS can make those existing AI algorithms more efficient at sifting through what would normally be a prohibitive number of potential actions allowing your game to do deeper analysis and, ultimately, provide better decisions for gameplay.
|