next up previous
Next: IPP Up: Our Experimental Setup Previous: Our Experimental Setup


The planners are all what have been called primitive-action planners [Wilkins desJardins2001], planners that require relatively limited domain knowledge and construct plans from simple action descriptions. Because the AIPS98 competition required planners to accept PDDL, the majority of planners used in this study were competition entrants or are later versions thereof 4. The common language facilitated comparison between the planners without having to address the effects of a translation step. The two exceptions were UCPOP and Prodigy; however, their representations are similar to PDDL and were translated automatically. The planners represent five different approaches to planning: plan graph analysis, planning as satisfiability, planning as heuristic search, state-space planning with learning and partial order planning. When possible, we used multiple versions of a planner, and not necessarily the most recent. Because we conducted this study over some period of time (almost 1.5 years), we froze the set early on; we are not comparing the performance to declare a winner and so did not think that the lack of recent versions undermined the results of testing our assumptions.


©2002 AI Access Foundation and Morgan Kaufmann Publishers. All rights reserved.