A quantitative theory for plan merging
AAAI'91 Proceedings of the ninth National conference on Artificial intelligence - Volume 2
| Hi-index | 0.00 |
Many realistic planning problems (such as those in manufacturing) place a high premium on plan efficiency. However, classical planning theory does not offer much insight into ways of obtaining efficient plans. The objective of this paper is to present a model of planning for plan efficiency that has been drawn from a case study in the machining domain. Some of the important features of this domain are that problems are stated as sets of conjunctive goals, and operators that achieve those goals have a high degree of overlap. Operators can be said to overlap when they can share work. Because of this overlap, the cost of the operators is dependent on their order in the plan, (for example, it is less time consuming to buy vegetables if your last action was also done at the grocery store) However, looking for a near optimal set of overlapping operators can lead a very expensive search. The methods that human machinists were found to be using reduced the complexity of the search for good operators by using cues and patterns in the problem specification, gained from experience, to tell them when it might be useful to explore an operator.