A Solution Method for Single-Source Capacitated Facility Location Problems Under Decision-Dependent Demand Uncertainty
Abstract
We present a model and an exact solution method for single-source capacitated facility location problems with decision-dependent demand uncertainty. The model allows the joint probability distribution of customer demands to depend on the facility to which demand points are allocated. The number of possible decision-dependent distributions is therefore equal to the number of feasible allocations of demand points to (groups of) facilities. To address the exponential size of the model, we apply and extend a recent version of the L-Shaped method for stochastic programs with decision-dependent uncertainty. In particular,we compare the original optimality cut with a new optimality cut based on a bilinear reformulation. We prove that, separated together with the original one, the new cut yields a tighter restriction, at the expense of introducing bilinear terms. In addition, we prove an effective decision-independent valid inequality that significantly improves convergence. Extensive numerical tests on 180 instances demonstrate that the method, supported by the valid inequality, can find optimal or near-optimal solutions to problem instances with up to 620 potential probability distributions. The tests also show that, while the new cut closes slightly more instances, the original cut remains preferable in terms of solution time and optimality gap.
Related
- P-curve: An improved solution for the single-facility location problem in two regions with ℓp- and ℓq-norms
- Volumes of the strong and weak relaxations of the uncapacitated facility-location polytope
- A Two-Stage Stochastic Facility Location Problem with Endogenous Demand Learning
- Enhanced Uncertainty Quantification Variational Autoencoders for the Solution of Bayesian Inverse Problems
- Fundamental solution expansions in stochastic homogenization of parabolic systems with time-dependent coefficients
BibTeX
@misc{pantuso2026solution,
title = {{A Solution Method for Single-Source Capacitated Facility Location Problems Under Decision-Dependent Demand Uncertainty}},
author = {Giovanni Pantuso and Miguel Lejeune},
howpublished = {arXiv},
year = {2026},
month = nov,
doi = {10.48550/arxiv.2611.08090},
eprint = {2611.08090},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2611.08090},
note = {CC BY-NC-SA 4.0},
}
RIS
TY - UNPB
TI - A Solution Method for Single-Source Capacitated Facility Location Problems Under Decision-Dependent Demand Uncertainty
AU - Giovanni Pantuso
AU - Miguel Lejeune
PY - 2026
DA - 2026/11/05
PB - arXiv
DO - 10.48550/arxiv.2611.08090
UR - https://arxiv.org/abs/2611.08090
AB - We present a model and an exact solution method for single-source capacitated facility location problems with decision-dependent demand uncertainty. The model allows the joint probability distribution of customer demands to depend on the facility to which demand points are allocated. The number of possible decision-dependent distributions is therefore equal to the number of feasible allocations of demand points to (groups of) facilities. To address the exponential size of the model, we apply and extend a recent version of the L-Shaped method for stochastic programs with decision-dependent uncertainty. In particular,we compare the original optimality cut with a new optimality cut based on a bilinear reformulation. We prove that, separated together with the original one, the new cut yields a tighter restriction, at the expense of introducing bilinear terms. In addition, we prove an effective decision-independent valid inequality that significantly improves convergence. Extensive numerical tests on 180 instances demonstrate that the method, supported by the valid inequality, can find optimal or near-optimal solutions to problem instances with up to 620 potential probability distributions. The tests also show that, while the new cut closes slightly more instances, the original cut remains preferable in terms of solution time and optimality gap.
ER -