A NOVEL CAUSAL GRAPH BASED HEURISTIC FOR SOLVING PLANNING PROBLEM
At present, heuristic planning is a focus in intelligence planning area. Two of the best known heuristic planners are HSP and FF. However, they are both based on the delete-relaxation, and some vital information of the planning task may be lost. Fast Downward is a domain-independent heuristic planner, which is based on the causal graph heuristic. Because of translating a problem to a multi-valued planning task, and exploiting the underlying causal structure, Fast Downward can yield better performance compared with the state of the art heuristic planners in many domains, but it needs the state variables to be independent, which is always not satisfied. Based on the causal graph heuristic, this paper proposes a novel causal graph mixed heuristic with two proportion coefficients, which is the combination of the additive heuristic and max heuristic. Experiments have proved that the search time and plan length can be reduced under the heuristic method, and we can make a tradeoff between them.
Causal graph mized heuristic Additive heuristic Maz heuristic Proportion coefficient
WEN-XIANG GU JIN-LI LI MING-HAO YIN JUN-SHU WANG JIN-YAN WANG
Department of Computer Science, School of Computer, Northeast Normal University, Changchun 130117, China
国际会议
2008 International Conference on Machine Learning and Cybernetics(2008机器学习与控制论国际会议)
昆明
英文
2223-2228
2008-07-12（万方平台首次上网日期，不代表论文的发表时间）