Lifter assignment algorithm of overhead hoist transport system in multi-floor semiconductor fab using deep learning

Abstract

This paper presents a lifter assignment problem in an overhead hoist transport (OHT) system in a multifloor semiconductor fab where each floor is connected via a lifter. Previous studies provided rule-based methods to consider only input port status and the expected transfer time. As the size of the fab has been increased, more factors should be considered when choosing a lifter, however, rule-based methods are limited in the increase of factors to be considered. In this work, we analyze and propose field-available methods by considering various factors affecting the lifter assignment. In this study, we propose the lifter assignment method that includes various factors that existing methods have not considered and verify performance improvement by using deep neural networks. Validation used Applied Materials’ AutoMod™ (version 14.0) simulation software.