Hierarchical path-finding based on decision tree

  • Authors:

  • Yan Li;Lan-Ming Su;Wen-Liang Li

  • Affiliations:

  • Machine Learning Center, Faculty of Mathematics and Computer Science, Hebei University, Baoding, China;Machine Learning Center, Faculty of Mathematics and Computer Science, Hebei University, Baoding, China;Machine Learning Center, Faculty of Mathematics and Computer Science, Hebei University, Baoding, China

  • Venue:
  • RSKT'12 Proceedings of the 7th international conference on Rough Sets and Knowledge Technology
  • Year:
  • 2012

Quantified Score

Hi-index 0.00

Visualization

Abstract

Path-finding is a fundamental problem in computer games, and its efficiency is mainly determined by the number of nodes it will expand. A* algorithm is unsuitable for path-finding on large map under limited computer sources and real-time demand, because the number of nodes it will expand grows fast with the size of the search space. HPA* can greatly improve the efficiency by generating abstract graph of the given map to memorize the map information before doing pathfinding. Through evenly partitioning the map as preprocessing, it can also reduce the influence of terrain factor on the output. As a result, it finds near optimal paths instead of optimal ones. And the evenly partition on the map doesnt consider the terrain distribution, which may still cause resource waste to some extent. In this paper, we present DT-HPA* (Hierarchical Path-Finding A* based on Decision Tree), a hierarchical path-finding approach on the map which has been divided by decision tree. This approach views each point on the map as an instance, and divides the map according to cut-points of continuous valued decision tree. The result of division is that the map is cut into some rectangular regions in different size, and retains the regions contain a kind of terrain. The experimental results show that, compared to HPA*, DT-HPA* can find more optimal paths with fewer detected nodes.