View article

In this paper we deal with the pathfinding problem in Real Time Strategy (RTS) games with partial observability for Artificial Intelligence (AI) agents. We first propose a Bootstrap JPS algorithm to perform pathfinding efficiently alongside a routine to preprocess terrain features. We then establish a grid system to learn map features systematically considering both mobile and immobile units. Utilizing the learned map features, we employ a waypoint-based pathfinding technique to navigate pathfinding agents away from threats efficiently. Using maps from a few established RTS games, we demonstrate the performance of our pathfinding framework and compare it with a few alternative approaches.