Models And Algorithms Of Uncertain Multilevel Programming

Multilevel programming was developed for dealing with decentralized decision-making problem, in which multiple decision makers arranged within a hierarchical ad-ministrative structure have their own decision variables, objectives and constraints. Theupper-level decision maker makes decision, then according to it, the lower-level de-cision makers can make decisions to optimize their own objectives. This dissertationstudies the decentralized decision system in uncertain environments and provides a gen-eral framework of uncertain multilevel programming with numerical solution methods.Decentralized decision systems in uncertain environments are very common andimportant in the area of economic systems, transportation, engineering, etc. Due to themodelling and computational difficulties, only an expected value model have been de-veloped for a simple stochastic two-level decision system. This dissertation deals withgeneral decentralized decision systems in stochastic and fuzzy environments. Firstly,stochastic expected value multilevel programming model, chance-constrained multi-level programming model and dependent-chance multilevel programming model areproposed, and a hybrid intelligent algorithm is designed for finding the Nash equi-librium and Stackelberg-Nash equilibrium. Secondly, three types of fuzzy multilevelprogramming models with concepts of Nash equilibrium and Stackelberg-Nash equi-librium are provided. For some special cases, the crisp equivalents are given, and forgeneral cases, a hybrid intelligent algorithm is designed by integrating fuzzy simu-lation, neural network and genetic algorithm. Lastly, an application of the proposeduncertain multilevel programming to the resource allocation problem is introduced.In conclusion, this dissertation brings forth a new and challenging topic: uncertainmultilevel programming including (1) stochastic multilevel programming models witha hybrid intelligent algorithm for the Nash equilibrium and Stackelberg-Nash equilib-rium; (2) fuzzy multilevel programming models, crisp equivalents, and a hybrid intel-ligent algorithm for finding Nash equilibrium and Stackelberg-Nash equilibrium; (3)applications of uncertain multilevel programming to the resource allocation problem.