AXIS: An interactive solution procedure combining structured and knowledge-based approaches to a multiple objective aggregate production planning problem

This dissertation develops and presents one method of combining highly structured algorithms that seek a "best" solution with the less structured information typically found in knowledge based expert systems. Both solutions approaches have drawbacks and advantages; AXIS combines them to minimize the drawbacks while maximizing the advantages. Specifically, solution time shows a substantial reduction while solution accuracy has only a slight decrease. AXIS is unique in that it is a truly interactive system, as opposed to the many combinations of these approaches that should more accurately be termed "sequential." The structured portion of AXIS is used to insure verifiable constraints are not violated and to search the resultant solution space. The unstructured portion of AXIS is used to guide the search of the structured portion, finding high quality solutions much faster than typical search procedures. Continually switching between the two approaches during the processing keeps AXIS from bogging down in a search of non-productive areas (a benefit of the intelligent search from the knowledge base) while all solutions generated are feasible (the benefit of the structured approach). The AXIS approach can be applied to any structured algorithm that has a generation/evaluation loop, and to any problem situation wherein heuristics are used for decision making. This dissertation applies the AXIS approach to a single algorithm (multiple objective integer programming) and a single problem setting (Aggregate Production Planning) to test the approach. The results from this prototype are sufficiently encouraging to warrant more detailed testing of this approach.