complicated task due to its broad scope and the strong connectedness of its objects and issues. In order to make theoretical investigations of supply chains feasible and to support decision-making in real world supply chains, simulation models are used. An integration of system dynamics and discrete agent-based modeling is a promising combination of methods for reducing the a priori complexity of the model. The paper discusses the strengths and weaknesses of system dynamics and discrete agent-based modeling. An approach for integration of the two modeling methodologies is presented. Issues concerning the practical coupling of software environments and a simple, prototypical supply chain model are discussed. Experiments for which the integrated simulation solution is applied are described. Insights in emergent structures in supply chains are derived from these simulation analyses.

This paper presents an approach to integrate the system dynamics and the agentbased modeling techniques. After reviewing the fundamental principles of the two modeling approaches, an agent-based supply chain simulation model is developed. The model consists of two levels of aggregation; on the macro level a discrete agent-based modeling approach is applied, on the micro, the agent level system dynamics is used to model the agents ’ internal cognitive structure. The paper concludes with first preliminary simulation results and aspects of future research. Coming from the field of complexity science, the agent-based modeling approach gains growing popularity. At the core of this perspective is the assumption that complexity arises from the interaction of individuals (Phelan, 2001). The behavior of these individuals, also called agents, is dictated by their schemata. According to Anderson, a schema is “a cognitive structure that determines what action the agent takes at time t, given its perception of the environment ” (Anderson, 1999). An agent’s schema can evolve over time what allows it to adapt to its environment. From a modeling perspective, this adaptation can be achieved by the use of feedback and learning algorithms (Phelan, 2001). In agent-based modeling schemata are mostly modeled as sets of simple generative rules. The above mention of structure and feedback already points at the use of system dynamics for modeling an agent’s schema. To evaluate the feasibility of this approach, the agent-based and the system dynamics modeling techniques have to be compared. Some of the basic differences between the two approaches are summarized in Table 1. Principle System Dynamics Agent-Based Modeling Building block Feedback loop connecting behavioral variables Individual agents connected by feedback loop Object of interest Structure of the system Agents ’ rules Research approach Development of object of interest over time Deductive: infer from structure to behavior Structure is fixed Inductive: infer from individual agents ’ behavior to system behavior Agents ’ rules can be adaptive