| Summary: | 博士 === 國立成功大學 === 工業與資訊管理學系 === 103 === This dissertation studies hierarchical optimization problems in a fuzzy environment. For two main decision structures of (decentralized) multi-level systems, we investigate decentralized bi-level multi-objective linear programming (DBL-MOLP) and multi-level linear programming (MLLP) problems with imprecise goals, and develop proper solution approaches based on fuzzy goal programming (FGP). For DBL-MOLP problems with multiple followers at the second level, decision decentralization is considered, and allowed to be adjusted depending upon the higher-level decision-maker (DM) to determine the linguistic levels of relative satisfaction of lower-level DMs compared to his own. For MLLP problems with more than one DM who can affect the feasible strategies of their subordinate-level DMs, we develop a two-phase FGP approach. In the first phase, a candidate solution that optimally satisfies the fuzzy goals of the DMs with the leader-follower requirement is suggested. In the second phase, a higher-level DM can adjust his satisfaction based on the level of relative satisfaction compared to that of his subordinate DM. Importantly, the proposed approaches ensure the leader-follower relationship in the solution processes by constraining the integrated (average) satisfaction of a DM to be not worse than that of his lower-level DMs based on the equal scales of fuzzy goals. The judgment in the adjustment processes, represented using linguistic terms, reflects the uncertainty in the problem. For greater efficiency, linearization is made for fractionally expressed constraints in the models. Through numerical examples, the feasibility and effectiveness of the proposed approaches are demonstrated.
|
|---|
