A robust mathematical model and ACO solution for multi-floor discrete layout problem with uncertain locations and demands

• Published in: Computers & industrial engineering Vol. 96; pp. 237 - 248
• Main Authors: Izadinia, Niloufar, Eshghi, Kourosh
• Format: Journal Article
• Language: English
• Published: New York Elsevier Ltd 01.06.2016; Pergamon Press Inc
• Subjects: Algorithms; Ant colony optimization; Buildings; Computer simulation; Constants; Industrial engineering; Integer programming; Materials handling; Mathematical analysis; Mathematical models; Mixed integer programming; Multi-floor layout problem; Optimization algorithms; Robust optimization; Stores; Studies; Uncertainty modeling; Ant colony optimization; Mixed integer programming; Multi-floor layout problem; Robust optimization; Uncertainty modeling
• ISSN: 0360-8352; 1879-0550
• DOI: 10.1016/j.cie.2016.02.026

•We model a new multi-floor layout problem.•Unequal-area departments should have no overlapping with each other and the elevator in a discrete environment.•We used two robust approaches for modeling, with consideration of uncertain locations, demands and material flows.•An ACO is developed for solving large-size instances. The Multi-Floor Layout Problem (MFLP) is the problem of finding the position of each department in a plant floor in a multi-floor building without any overlapping between departments in order to optimize a particular objective function, more commonly the sum of the material handling costs. In this paper, a special class of MFLP, called Uncertain Multi-Floor Discrete Layout Problem (UMFDLP), is defined. In this problem, a multi-floor building is considered in which an underground store is utilized to contain main storages, and different departments can be located in the other floors in potential pre-determined locations. Furthermore, all material flows are not constant. Moreover, the locations for departments can be chosen from intervals, where no overlaps are allowed. We develop a Mixed Integer Programming (MIP) model to generate a robust solution for UMFDLP. Furthermore, the lower bound of objective function is obtained. Moreover, an ACO algorithm is designed for solving large instances. Then, a set of problems is generated and tested by the proposed algorithm. The results show the efficiency of our model and algorithm.