Relative Utility of EBG based Plan Reuse in Partial Ordering vs. Total Ordering Planning

This paper provides a systematic analysis of the relative utility of basing EBG based plan reuse techniques in partial ordering vs. total ordering planning frameworks. We separate the potential advantages into those related to storage compaction, and those related to the ability to exploit stored plans. We observe that the storage compactions provided by partially ordered partially instantiated plans can, to a large ex1ent, be exploited regardless of the underlying planner. We argue that it is in the ability to exploit stored plans during planning that partial ordering planners have some distinct advantages. In particular, to be able to flexibly reuse and extend the retrieved plans, a planner needs the ability to arbitrarily and efficiently "splice in" new steps and sub-plans into the retrieved plan. This is where partial ordering planners, with their least-commitment strategy, and flexible plan representations, score significantly over state-based planners as well as planners that search in the space of totally ordered plans. We will clarify and supporll this hypothesis through an empirical study of three planners and two reuse strategies.

[1]  Richard E. Korf,et al.  Planning as Search: A Quantitative Approach , 1987, Artif. Intell..

[2]  Oren Etzioni,et al.  Statistical methods for analyzing speedup learning experiments , 2004, Machine Learning.

[3]  Manuela Veloso Learning by analogical reasoning in general problem-solving , 1992 .

[4]  Oren Etzioni,et al.  Explanation-Based Learning: A Problem Solving Perspective , 1989, Artif. Intell..

[5]  David E. Smith,et al.  A critical look at Koblock's hierarchy mechanism , 1992 .

[6]  Anthony Barrett,et al.  Partial-Order Planning: Evaluating Possible Efficiency Gains , 1994, Artificial Intelligence.

[7]  James A. Hendler,et al.  A Validation-Structure-Based Theory of Plan Modification and Reuse , 1992, Artif. Intell..

[8]  Subbarao Kambhampati,et al.  A Unified Framework for Explanation-Based Generalization of Partially Ordered and Partially Instantiated Plans , 1994, Artif. Intell..

[9]  Richard Fikes,et al.  Learning and Executing Generalized Robot Plans , 1993, Artif. Intell..

[10]  Craig A. Knoblock Learning Abstraction Hierarchies for Problem Solving , 1990, AAAI.

[11]  Drew McDermott,et al.  Regression planning , 1991, Int. J. Intell. Syst..

[12]  M. Veloso,et al.  Nonlinear Planning with Parallel Resource Allocation , 1990 .

[13]  Subbarao Kambhampati,et al.  EXPLOITING CAUSAL STRUCTURE TO CONTROL RETRIEVAL AND REFITTING DURING PLAN REUSE , 1994, Comput. Intell..

[14]  Stan Matwin,et al.  Knowledge and regularity in planning , 1992 .

[15]  Steven Minton,et al.  Total Order vs. Partial Order Planning: Factors Influencing Performance , 1992, KR.

[16]  David A. McAllester,et al.  Systematic Nonlinear Planning , 1991, AAAI.