A Framework for Case - Based Revision for Schedule Generation and Reactive Schedule Management

概要

論文の詳細を見る
We describe a unified framework for job shop schedule optimization in predictive and reactive contexts by iterative revision. The approach elicits user's context-dependent scheduling preferences in the form of cases through interaction with a user, and re-uses them to dynamically guide schedule revision decisions. The goals of the approach are : (1) correctness of the resulting schedule in the sense that there are no violated constraints, (2) optimization of the modified schedule according to the user's input criteria and preferences, and (3) minimization of schedule disruptions in the reactive contexts. The methodology integrates case-based reasoning (CBR) mechanisms for incremental accumulation and re-use of past repair experiences that incorporate user preferences and schedule evaluation explanations, and constraint-based scheduling for propagation and resolution of the effects of repair. This framework is implemented in a system called CABINS. Experimental evaluation of CABINS as compared with other scheduling methods shows that the approach is able to (a) capture and effectively utilize user's scheduling preferences that were not present in the scheduling model, (b) outperforms other scheduling methods in both predictive schedule generation and reactive response to unpredictable execution time events along a variety of criteria that have been recognized as crucial in real operating environments.
社団法人人工知能学会の論文
1994-05-01