Automatic planning and flexible scheduling: A knowledge-based approach

Planning in flexible manufacturing systems must take into account both the multiprocessing environment and the dynamically changing states. This paper uses a knowledge-based approach to handle such a planning system, where three levels-i.e., the control, state, and operational levels-constitute its manufacturing knowledge. The construction of plans for multiple manufacturing jobs requires four steps: (1) linear plan generation, (2) conflict detection, (3) plan synthesis, and (4) plan revision. To achieve goals, these steps will search for planning operators using a backward chaining inference procedure. To perform the scheduling and sequencing of the multiple jobs within a flexible manufacturing cell, the planning system modifies the nonlinear planning approach and adds resources and durations information to the action formalism. Because the plan-generation process is goal-directed, the dynamically adjustable plan can be constructed on-line.