Research On Modeling And Simulation Of Distributed Inventory System Based On Multi-Agent System

An inventory control policy in a distributed inventory system (DIS) with local decision transshipment policy is studied in the thesis. Considering in a two-echelon supply chain with one supplier and multiple retailers, multiple retailers form a DIS in which continuous review and oder-up-to S policy is applied. Supplier monitores the inventory status of each retailer and takes a central replenishment when the whole inventory of DIS is lower than s. When on-hand inventory and pipline stock are unavailable for demand, retailers have to cooperate with each othe for inventory transshipments to avoid back off. Consultations between retailers about transshipment are the processes of local decision. A case of a two-echelon supply chain consisting of six retailers and one common supplier is taken for example. Through simulation and optimization based on simulation experiments, the optimal inventory control policy and rules of DIS are concluded. The following sides are mainly studied:1) Based on the systematic summary about the former relative studies of DIS problem, DIS's definition, realm and content are concluded.2) Based on MAS theory, a hierarchical model of DIS with the feather of local decision is built up. UML is used to describe the coorperation and consultation between multiple Agents.3) The transshipment consultation of DIS is considered as an unlimited repeated game and Robert Axelrod's evolution theory of cooperation is brought into the transshipment cooperation. "Tit for Tat" policy which means the cooperation of last time would influence that of the next time, which is a development for transshipment research. Contract Net protocol mechanism is applied to the transshipment consultations. Spontaneous organization and local decision which kept transshipment makes every retailer get along orderly.4) Central control in replenishment and local decision in transshipment policy are combined in the model of DIS. The influence that transshipment bringing into inventory control policy makes the DIS has its own feathers. The result of our study indicates that DIS has some rules and features of itself. 5) This paper built a model based on reality. Many necessary factors of system are considered, such as various possible lead-times, different parameters of every retailer, complex communication and so on. The DIS model is supported by the simulation on AnyLogic very well. Some new acknowledgement about features and rules of DIS is found from excessive simulation and optimization experiments.