Evaluating Adaptive Customer Strategies in TAC SCMYevgeniya Kovalchuk, Maria FasliDepartment of Computing and Electronic Systems, University of EssexWivenhoe Park, Colchester, CO4 3SQ, UK{yvkova; mfasli}@essex.ac.ukAbstractSupply Chain Management (SCM) is a complex processwhich includes a number of interrelated activities such as:negotiating with suppliers for raw materials, competing forcustomer orders, managing inventory, schedulingproduction, and delivering goods to customers. In thispaper we present a number of strategies to be examined inthe domain of SCM. We introduce a multi-agent systemwhich we used to evaluate the proposed methods. Wetested the system in the Trading Agent Competition SCMgame, which offers a realistic simulated environment forstudying SCM strategies. Although we introduce a numberof strategies, we concentrate on the ones for predictingwinning bidding customer prices to support a successfulperformance on the customer side of the supply chain.1 IntroductionIn today's highly dynamic, time-constrainedenvironments, developing efficient decision supportsystems is a key challenge. In particular, in the domain ofSCM, which deals with the planning and coordination ofthe activities of organizations from getting raw materials,manufacturing goods to delivering them to customers,supporting dynamic strategies is a major but unresolvedissue. All entities in the supply chain are highly connectedand interdependent. Being successful in one area of thesupply chain does not necessarily guarantee theimprovement of the overall performance. Thus, there isthe need for a mechanism to separate different tasks andexplore them both independently and in relation to eachother. We implemented such a mechanism in our multiagentdecision support system which we tested in theTAC SCM game (Collins et al. 2006). Using a multiagentapproach, we built a number of TAC SCM agentsand allowed them to compete against each other in orderto compare the performance of each proposed algorithm.
đang được dịch, vui lòng đợi..
