.png){kind=icon}
.png){kind=icon}
Return
Three-dimensional facility layout optimization using large language model-driven A-star and non-dominated sorting genetic Algorithm-II
.png){kind=icon}
PRE
AI
DOI：10.1016/j.eswa.2026.131745.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Facility layout optimization is critical for reducing logistics costs, delivery time, and land use. With increasing land resource constraints, facility layouts are shifting from two-dimensional to three-dimensional configurations, invoking vertical transport paths between multiple floors. However, existing multi-floor facility layout studies typically optimize the layout first and then expand the facility area to accommodate logistics channels, often neglecting the integration between the two, which results in suboptimal layout outcomes. This paper proposes a syncing optimization model that integrates three-dimensional facility layout with flexible logistics channel design. To construct this model, this paper focuses on the often-overlooked actual cost function in A-star algorithm and leverage large language model to generate a multi-objective-aware evaluation function for path planning. Large language model acts as an online controller to dynamically adjust the crossover and mutation probabilities in non-dominated sorting genetic algorithm-II, enabling adaptive evolution in response to the high complexity arising from numerous facilities and interrelated constraints. A case study on a prefabricated component production line shows that, compared to classic path-planning algorithms, only the currently proposed large language model-driven A-star algorithm is able to find the shortest paths while minimizing the number of turns across 14 layouts. Additionally, large language model driven non-dominated sorting genetic algorithm-II outperforms the baseline algorithm, with 60.78% reduction in average inverted generational distance and 55.64% improvement in average hypervolume. Compared to traditional layout results, the layout scheme using the syncing optimization model reduces logistics costs by 29%, shortens logistics time by 25%, and increases space-use rate by 19%.
Keywords:
Three-dimensional facility layout
Logistics channel design
Large language model
A-star algorithm
Non-dominated sorting genetic algorithm-II
Journal
IF:
7.5
Papers:
2.9W
Citations:
10.2W