The Study On Optimization Method For Sequential Decision-making In Supply Chain

The sequential decision-making in the supply chain is a process, in which deci-sion makers may encounter some new situations after making a decision, and they should make the rest decisions according to new situations. Due to the information being revealed, the dependency relationship between the present event and previous event, the decision makers not only need to evaluate the current situation, but also have to trade off the advantages and disadvantages of the future events. There are many uncertain factors during a decision-making process. If the decision makers make their decisions without updating information, it not only will not obtain the maximum profit, but also bring huge risk to the supply chain. The purpose of the sequential decision-making is to reduce the risk in the supply chain caused by vari-ous uncertainties as much as possible. Using the market feedback information and making appropriate sequential decisions are the key research issues in this thesis.Because the uncertainty in the sequential decision-making process may come from the objective or subjective factors, the developed work in this thesis is to study the sequential decision problems in the supply chain based on the stochastic and fuzzy theory. In stochastic environment, three typically sequential decision-making problems are studied. Two-level stochastic expected value models are established for the sequential decision-making problems. A new solution method is proposed; it first handles the lower level programming problem, and then solves the upper level programming problem to obtain the optimal decisions. Some numerical experiments are proposed to demonstrate the proposed modeling idea and the feasibility of solu-tion method. In fuzzy environment, the simultaneous decision problem in the supply chain is studied, and the equivalent sub-models of the original model are obtained. Secondly, two kinds of sequential decision problems are discussed. The profits are measured by equivalent value criterion, two-level equivalent value optimization mod-els are established and their equivalent deterministic programming models are solved via optimization software. Finally, some numerical experiments are conducted to demonstrate the proposed modeling idea and the feasibility of solution method.The major new results of the thesis include the following several aspects:(i) In stochastic environment, when price decision is before the quantity decision, a new two-level expected value model is established for the sequence decision problem. A general solution to the lower level programming problem is obtained, (ii) When both yield and demand follow uniform distributions, the analytical optimal quantity decision are obtained, and the equation that the optimal price decision satisfies is obtained. (iii) When quantity decision is before price decision, two new two-level stochastic expected value models are established for two kinds of sequential decision problems. (iv) For the stochastic sequential decision problem that price decision before quantity decision, some numerical experiments are conducted and sensitivity analyses for model parameters are provided. (v) In fuzzy environment, three kinds of sequential decision problems are studied, they are simultaneous decision making, quantity decision before price decision and price decision before quantity decision. Three new equivalent value optimization models are established, and their equiva-lent deterministic programming models are solved via optimization software. Some numerical experiments are conducted to demonstrate the proposed new modeling idea and the feasibility of the designed solution methods.