吉林大学学报(理学版)

• 计算机科学 • 上一篇    下一篇

结合Separator的约束满足问题树分解方法

吕巍, 张舒娟   

  1. 吉林大学 计算机科学与技术学院, 长春 130012
  • 收稿日期:2015-04-21 出版日期:2016-03-26 发布日期:2016-03-23
  • 通讯作者: 张舒娟 E-mail:zhangshujuan2009@yeah.net

Tree Decomposition Method Combined with Separator in Constraint Satisfaction Problems

LV Wei, ZHANG Shujuan   

  1. College of Computer Science and Technology, Jilin University, Changchun 130012, China
  • Received:2015-04-21 Online:2016-03-26 Published:2016-03-23
  • Contact: ZHANG Shujuan E-mail:zhangshujuan2009@yeah.net

摘要:

基于树分解的回溯搜索算法, 结合separator分解算子提出一种新的搜索算法BTD+-MAC. 该算法在搜索时, 优先选择separator中的变量进行相容性检查和实例化, 由于树宽度的减小能提高约束传播的效率, 进而提高问题求解效率. 对几组benchmark问题进行测试, 测试结果表明, 该算法在问题求解效率上超过了MAC3rm算法和BTD-MAC算法.

关键词: 人工智能, 约束满足问题, 树分解, 回溯搜索, 启发式方法

Abstract:

Based on the tree decomposition backtracking search algorithm, combined with the separator decomposition operator, we proposed a new search algorithm BTD+-MAC. The algorithm first selected the variables in separator to carry on the consistency checks and instantiation during search, so the efficiency of constraint propagation could be improved by reducing the width of the tree and further improved the efficiency of solving problems. We tested several sets of benchmark problems. The results show that the algorithm is more efficient than the MAC3rm and BTD-MAC algorithm in solving the problem.

Key words: artificial intelligence, constraint satisfaction problem, tree decomposition, backtracking search, heuristic method

中图分类号: 

  • TP18