Asymmetric Action Abstractions for Multi-Unit Control in Adversarial Real-Time Games
This addresses the challenge of real-time strategy game AI for developers and researchers, offering a novel abstraction method to improve performance in large-scale scenarios.
The paper tackles the problem of deriving effective strategies in large-scale multi-unit real-time adversarial games by introducing asymmetric action abstractions, which retain theoretical advantages over regular abstractions while allowing search algorithms to outperform state-of-the-art approaches in combat scenarios.
Action abstractions restrict the number of legal actions available during search in multi-unit real-time adversarial games, thus allowing algorithms to focus their search on a set of promising actions. Optimal strategies derived from un-abstracted spaces are guaranteed to be no worse than optimal strategies derived from action-abstracted spaces. In practice, however, due to real-time constraints and the state space size, one is only able to derive good strategies in un-abstracted spaces in small-scale games. In this paper we introduce search algorithms that use an action abstraction scheme we call asymmetric abstraction. Asymmetric abstractions retain the un-abstracted spaces' theoretical advantage over regularly abstracted spaces while still allowing the search algorithms to derive effective strategies, even in large-scale games. Empirical results on combat scenarios that arise in a real-time strategy game show that our search algorithms are able to substantially outperform state-of-the-art approaches.