A Novel Space Reduction Algorithm for Planning

Most automated planners generate plans using heuristics search. However, they still face scalability challenges for large-scale problems. Space reduction has always been an attractive idea to planning researchers. We propose a novel algorithm to reduce the search space using knowledge tree structure, whose nodes are actions to achieve a predicate. Based on the structure, we only expand a subset of relevant actions at each state during planning, without searching other useless actions. We test our algorithm in several planning domains in the experiment, and the experiment results show that our algorithm is efficient on space reduction.