Li Tan, Shenghan Xu, Benjamin Meyer, and Brock Erwin. In International Journal of Information and Decision Sciences. Vol. 4, Nos. 2/3, InderScience. 2012.
We propose an extensible object-oriented agent-based framework for modeling and simulating supply chains. In today's economy many companies rely on diversified global supply chains to reduce cost and attain competitive edge. The structures and behaviors of these supply chains vary based on underlying business models and markets. However, most of existing supply-chain analysis tools are designed only for specific subsets of supply chains. The primary goal of this work is to provide an open and extensible framework for analyzing supply chains with heterogenous elements and network structures. To improve extensibility, the framework has incorporated the following features: 1) the framework adopts an agent-based approach to handle interactions among elements of a supply chain: elements are modeled as autonomous agents and their interactions decide the behavior of the supply chain. The framework provides common functionalities for studying interactions among elements, and an analyst has freedom to define new types of elements required for a specific supply-chain application; and, 2) to improve the design reusability and ease the difficulty in defining a new type of element, we propose an object-oriented type system that supports behavior inheritance. An analyst can focus on defining functions unique to a new type of element, and inherit common behaviors from existing types; and, 3) the framework includes a meta-model for elements of a supply chain. The meta-model abstracts the interface and behavior of an element from its role. We define the formal semantics of the meta-model. The formal semantics helps remove ambiguity in defining a supply-chain model and interpreting analysis result; and, 4) the framework includes a discrete-event simulation algorithm. While the meta-model defines the interface and behavior of individual element, the simulation algorithm defines interactions among elements of a supply chain via messages and deliveries. We will also discuss SimRisk, our Java implementation of the proposed framework.