Schedule-Informed Empty-Vehicle Dispatching for OHT Systems in Semiconductor Manufacturing

Abstract

Overhead hoist transport (OHT) performance affects the execution of semiconductor production schedules because transport delays postpone subsequent operations. This study develops the Scheduler and OHT Allocation Monitoring Simulator (SOMOS), an event-driven environment that integrates machine processing, transport request generation, OHT movement, and subsequent production execution. Based on SOMOS, we propose Schedule-informed Empty-Vehicle Dispatching (SEVD), which repositions idle OHT vehicles before pickup requests occur using immediate pickup demand, remaining processing information, and a schedule-derived spatial demand prior. SEVD retains the conventional HUNGARIAN assignment procedure for actual requests. Experiments compare SEVD with six benchmark dispatching methods over 20 paired seeds. Relative to HUNGARIAN assignment, SEVD reduces the mean post-assignment empty-approach travel from 96.82 s to 71.23 s and source wait p90 from 428.74 s to 399.34 s while maintaining production-level performance comparable to the strongest benchmarks. These gains are obtained by redistributing vehicle movement toward proactive repositioning, with total empty travel increasing by 17.8% and total travel distance by 6.8%. Fleet size and ablation experiments further show that the clearest benefit is associated with proactive spatial repositioning rather than precise temporal prediction. These results support interpreting SEVD as a framework for the proactive spatial repositioning of idle vehicles, while the incremental contribution of schedule-derived information was not statistically detectable in the present testbed.

The authors' abstract, as published at the source. Systems, 2026 · DOI ↗