.png){kind=icon}
Return
An Adaptive Sampling Method for Parallel Simulation-based Optimization in Automated Material Handling Systems
.png){kind=icon}
PRE
AI
DOI：10.1016/j.simpat.2026.103354.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
An automated material handling system (AMHS) in semiconductor and display fabrication facilities is crucial for enhancing productivity by managing the material flow across numerous production steps via vehicles. AMHS vehicle management is typically formulated as a simulation-based optimization (SBO) problem due to complex operational logic and stochastic system dynamics. However, practical SBO is computationally intensive because reliable performance estimation requires multiple replications per candidate solution, and exhaustive exploration is infeasible under a limited computational budget. To address these challenges, this study proposes a parallel SBO framework with adaptive sampling. The framework employs decision tree-based adaptive sampling (DTAS) to focus the search on high-potential regions and allocates replications dynamically according to output variability. Moreover, parallel evaluation with replication rebalancing is incorporated to fully utilize available computing resources and reduce overall computation time. Simulation experiments show that the proposed algorithm reduces delivery time and computation time by 7.05% and 35.71%, respectively, in the idle vehicle repositioning problem and by 4.24% and 15.53%, respectively, in the vehicle design problem, relative to benchmark algorithms.
Keywords:
Automated material handling system
Simulation-based optimization
Adaptive sampling
Parallel computing
Decision tree
Journal
IF:
4.6
Papers:
2.6K
Citations:
4.8K
Organization
Cited Papers
No cited papers available