Benders decomposition for congested partial set covering location with uncertain demand

TL;DR

This paper develops a Benders decomposition approach for a robust, quadratic mixed-integer formulation of a congested partial set covering location problem under demand uncertainty, improving solution efficiency.

Contribution

It introduces a novel Benders decomposition method tailored for a robust quadratic mixed-integer problem, enhancing scalability and solution quality for large instances.

Findings

Benders approach outperforms state-of-the-art solvers on benchmark instances.

Robustness level significantly affects planning sensitivity.

Proposed perturbation technique improves Benders subproblem stability.

Abstract

In this paper, we introduce a mixed integer quadratic formulation for the congested variant of the partial set covering location problem, which involves determining a subset of facility locations to open and efficiently allocating customers to these facilities to minimize the combined costs of facility opening and congestion while ensuring target coverage. To enhance the resilience of the solution against demand fluctuations, we address the case under uncertain customer demand using $\Gamma$-robustness. We formulate the deterministic problem and its robust counterpart as mixed-integer quadratic problems. We investigate the effect of the protection level in adapted instances from the literature to provide critical insights into how sensitive the planning is to the protection level. Moreover, since the size of the robust counterpart grows with the number of customers, which could be significant in real-world contexts, we propose the use of Benders decomposition to effectively reduce the number of variables by projecting out of the master problem all the variables dependent on the number of customers. We illustrate how to incorporate our Benders approach within a mixed-integer second-order cone programming (MISOCP) solver, addressing explicitly all the ingredients that are instrumental for its success. We discuss single-tree and multi-tree approaches and introduce a perturbation technique to deal with the degeneracy of the Benders subproblem efficiently. Our tailored Benders approaches outperform the perspective reformulation solved using the state-of-the-art MISOCP solver Gurobi on adapted instances from the literature.