Akey question in conditional planning is: howmany, and whichof the possible executionfailures shouldbe plannedfor? Onecannot, in general, plan for all the possiblefailures becausethe searchspaceis too large. Onecannotignoreall the possible failures, or onewill fail to producesufficiently flexibleplans. In this paper, wedescribe an approachto conditional planning that attempts to identify the contingenciesthat contribute the mostto a plan’s overall utility. Plan generation proceeds by handling the most important contingencies first, extendingthe plan to include actions that will be taken in case the contingencyfails. Wediscuss the representational issues that mustbe addressedin order to implementsuch an algorithm, and present an examplewhichillustrates our approach.
Nilufer Onder, Martha E. Pollack