Analysis of Conventional Robot Task Sequences During Lot Switching Periods in Cluster Tools

摘要

The semiconductor manufacturing system operates in lot units comprising wafers with identical recipes. Production predominantly occurs within a cluster tool, encompassing multiple single-wafer processing chambers and a wafer-handling robot. Recently, the increasing popularity of small lot sizes has been driven by the diversification of customer demand and larger wafer diameters. Thus, cluster tools frequently encounter lot switching periods, where lots with different recipes are processed consecutively. Existing research on cluster tool scheduling primarily relies on conventional backward and swap sequences optimized for cyclic robot operations. This approach persists even during lot switching period scheduling, which represents non-cyclic scenarios. However, it remains unclear whether these conventional robot task sequences remain optimal or ensure consistent performance during lot switching periods. Therefore, we introduce a theoretical analysis of the performance of the conventional robot sequences, such as backward and swap sequences, in lot switching periods for the first time. We examine various dominant relations, establishing the conditions under which the conventional sequences provide an optimal schedule. We also show worst-case performance bounds. Finally, given the limited performance of the conventional swap sequence during lot switching periods for dual-armed robots, we propose a new form of the robot task sequence and demonstrate its effectiveness.