From: AAAI Technical Report SS-93-03. Compilation copyright © 1993, AAAI (www.aaai.org). All rights reserved. Integrating Planning and Scheduling Nicola Muscettola Centerfor Integrated Manufacturing Decision Systems TheRoboticsInstitute CarnegieMellonUniversity Pittsburgh, PA15213 Whenaddressingspecific classes of problems,it is not sufficient for a problemsolving paradigmto enable domaindescriptions and solution methods; it is necessaryto supportthe processas well. This requires the provisionof facilities that help makeit reasonably easy, free fromerrors andefficient. Weare interested in generating detailed predictions of the behaviorof large scale systems; such problems arise in the operations managementof complexsystems, such as space-basedfacilities andtransportationsystems.Here, the mainchallengeis the integration of planningand scheduling. Classical planning does not provide reasonable supportto the solution of these problems.In essence,it encouragesa viewof the world wherechange(action) operates as a labeled transition betweentwodifferent world states, persisting for undeterminedperiods of time. Thestructural complexityof a state is unbound but, paradoxically, the devices provided for its description are unstructured such as completefirst order theories or lists of predicates. This lack of balance betweengenerality and structuredness is a major obstacle whenaddressing integrated planning and schedulingproblems.Past research has identified several other problems:representation of processes continuouslyevolvingover time, parallelism, external events, etc. Althoughsolutions for individual aspects have been proposed, no comprehensiveapproach is currentlycapableof addressingthemall. Thelack of an integrated frameworkfor the solution of prediction problemshas hadtwo mainconsequences. First, the present complete separation between planning and scheduling competenceshas limited the rangeof solution methodologies available to us. In the classical approach, a pure (process) planning phase precedes a pure scheduling phase. However,it is apparent that this macroseparation cannot be easily advocated in several complex domains. While scheduling,for example,it is often necessaryto expand setup activities; these are not justified by the achievementof a goal but dependexclusively on how other activities are sequenced on a resource. Moreover,delayingplanninginto the schedulingphase mightallow the selection of courses of action that, althougha priori sub-optimal,are clearly convenient 162 whenconsidering the expectedusage of resources over time. Second, the difficulty in generating extendeded detailed predictions has encouragedthe attempt to completely avoid the process. In approaches that advocatethis viewclassical planningoperators can be readily used as simple stimulus-responserules. The rationale behind the "denial of prediction" is that since the conditions at execution are usually unpredictably different than those assumedduring planning, it is futile to look any further than the immediate future. Although unpredictability and reactivity mustbe certainly takeninto account,a purely reactive approachis not acceptable in domainswith complextimingconstraints and optimizationconcerns. Temporalmyopia,for example,mightgenerate gluts of resourcerequests in the future. Theconsequence would be either a suboptimalsystem utilization or actual systemfailure. Ourresearch is addressingall the previousissues in a integrated planning and scheduling framework[1]. The representational paradigm stresses the decomposition of a domain as a vector of state variables continuouslyevolvingover time. Eachstate variable is in effect a generalization of a classical scheduling resource; beyondits availability it also describes its state at any point in time. Temporal flexibility in the final scheduleallowsa certain amount of adjustmentduring execution, to adapt the extended prediction to unpredictable events. Wehave also demonstrated the advantages of coupling temporal flexibility with statistical estimates of expected resource congestionduringthe generation of extended predictions[2]. [1] Muscettola, N., Smith, S.F., Cesta, A., D’Aloisi, D. CoordinatingSpaceTelescopeOperationsin an Integrated Planningand Scheduling Architecture. IEEEControl SystemsMagazine12(2), 1992. [2] Muscettola,N. Schedulingby Iterative Partition of Bottleneck Conflicts. In Proceedings ofCAIA-93. IEEEComputer SocietyPress, 1993.