An agent-based infrastructure for assessing the performance of planning approaches for semiconductor supply chains

Abstract

Semiconductor supply chains that manufacture integrated circuits consist of dozens of semiconductor wafer fabrication, assembly and testing and storage facilities. The sheer size of the geographically distributed facilities and the involved supply chains, the pervasive presence of different kinds of uncertainties and the rapid pace of change lead to large-scale, distributed planning problems. However, the interactions between the different planning functions in semiconductor supply chains are not well understood. Systematic study of the interplay between the various planning algorithms in a real-world setting is extremely difficult due to the different software systems and decision-makers involved. In this paper, we discuss the design and implementation of a software agent-based experimental infrastructure for assessing the performance of different planning functions in a risk-free simulation environment, using web services to implement parts of the planning functionality. The environment is highly scalable since the planning functionality is provided by cloud computing mechanisms. Discrete-event simulation is used to emulate the underlying supply chain and execute the planning functions in a rolling horizon setting. We illustrate the benefits of the experimental environment and the benefits of using distributed computing techniques by studying the interaction between master planning and demand fulfillment, and between master planning and production planning in simplified semiconductor supply chains.