The development and validation of agent-based communication and negotiation models for apparel supply chain management

Abstract

Agent-based technology, a new approach for managing the supply chain at the tactical and operational levels, has emerged in recent years. It views a supply chain as one composed of a series of intelligent (software) agents, who are responsible for one or more activities and interacting with other related agents in planning and executing their expected goals. The agent-based technology has many beneficial features for autonomous, collaborative, and intelligent systems in distributed environments, which make it one of the best candidates for managing complex supply chains. Increasing pressures from customers for prompt deliveries and greater product variety at lower prices have impacted all entities of the supply chain -----the net work of organizations from raw material suppliers to manufacturers and distributors of finished goods. In response, manufacturing firms are moving towards collaborative fulfillment, partnering closely with other firms both upstream and downstream in the apparel supply chain to meet market needs. As these strategic alliances grow, firms have come to recognize the importance of proper negotiation to establish the terms of collaboration and interaction with partners, and of effective coordination activities and decisions, in lowering costs and in improving service. Automating negotiation as much as possible is a critical and challenging research problem. Solving this problem can benefit several areas of interest: electronic commerce, supply chain management, manufacturing resource planning and scheduling, distributed artificial intelligence, and multi-agent collaborative systems. This thesis presents methodologies and techniques needed for conducting automated negotiations along the apparel supply chain. The thesis identifies the problem of incorrect communication among supply chain partners, the unbalance between production capacity and the product due date and the effects of bid/offer price on supply chain cost. To illustrate the decision process, the problems that cover three stages-----source, make and delivery-----of the apparel supply chain, featuring inter-functional relationship within customer, merchandiser, manufacturer and supplier are presented. After reviewing the literature about the supply chain management, the relevant models and the application of agent technology in supply chain, in this research, the potential use of agent technology in the apparel supply chain has been explored. The knowledge flow through the supply chain processes has been depicted by unified modeling language (UML). The general research methodology adopted for investigating the application of agent technology in decision making in the negotiation process in the apparel supply chain has been specified. In this research, agent technology is used to facilitate decision making in negotiation procedure among merchandisers, manufacturers, suppliers and customers so as to enhance the coordination in the apparel supply chain. A simulation model based on Java platform is developed to emulate the interaction and communication of agents representing different partners in the apparel supply chain. The system describes the overall communication of agents in the apparel supply chain and represents a range of supply chain situations. To specify the negotiation process, another simulation model based on the software ZEUS® is built to display the negotiation procedure. Based on the agent-based negotiation model, an optimized negotiation model has been developed. To study and compare different negotiation strategies, the optimized negotiation model is proposed to have two negotiators involved, namely the manufacture agent and the supplier agent. Delivery date and price are considered as two major negotiation issues in this model. Three negotiation strategies with artificial algorithms are embedded into the optimization model. The first is the linear method which is considered as the basic operating method. The second is a two-phase negotiation strategy based on simulated annealing (SA) and genetic algorithm (GA) which has been developed to decrease the production cost and achieve better profit. The last is a genetic simulated annealing algorithm (GSAA) which integrates the advantage of GA and SA has been proposed to optimize the searching method to find a more accurate point. These algorithms are embedded into the optimized simulation model to improve the negotiation performance. In addition, a mathematical decision model have been built to support the decision requirements of the determination on price and delivery date and to assist obtaining an optimal due date and bargain price so that mutual profit can be maximized and supply chain cost minimized. Evaluation was achieved by comparison of different negotiation strategies adopting SA, GA and GSAA so that the optimal and ideal result can be verified. Industrial data was utilized to prove the feasibility of the agent-based negotiation model. For this purpose, a simulation-based optimization model for negotiation strategy in the apparel supply chain was proposed. It considers both supply chain cost mainly from production cost and inventory cost and the constraints of the manufacturers' production capacity. This simulation based-optimization model consists of two main parts, namely, the model for simulating the supply chain negotiation and the optimization model with intelligent algorithms for searching optimal negotiation strategies. All the models identify new decision problems in collaborative supply chains and provide novel solution approaches to solve the decision problems in negotiation effectively. On the whole, the simulation model was developed to explore the integration of artificial intelligence with human negotiation, probe on the improvement of negotiation strategy with optimized algorithms and study the effect of negotiation strategy (price, delivery date) on supply chain cost, mutual profit and production balance in an agent-based apparel supply chain.