Research and Development of StepByStep AI Planner

Abstract: AI planner is one of the important representations of AI planning study, the performances of planner, the efficiency and quality of plan, represent the researches of AI planning directly. This paper introduces the architectures of AI planner and StepByStep planner in brief, then describes the methods and strategies adopted by SteByStep in detail, defines the knowledge tree of predicate for extracting domain knowledge. Based on knowledge tree of predicate, the planning tree of predicate is defined and some strategies are applied for constructing the planning tree fast. StepByStep adopts some planning strategies according to the planning trees. Finally, some experiments are taken for eight planners with three representative benchmark domains and their problems, the performances of StepByStep, the efficiency and quality of plan, are analyzed in detail. Experiments show that the strategies of StepByStep controls the process of planning the problems of the three domains well, and validate th