Graph-based decision support system for last-mile humanitarian logistics: The case of Peru

Machuca L.; Briceño-Egúsquiza M.; Quiliche R.; Luna A.; Lazo J.; Chong M. · Issues in Information Systems (2026) · DOI: 10.48009/4_iis_2026_140

Resumen

This study demonstrates that existing social infrastructure can be strategically repurposed into safe clustering and distribution hubs during extreme crises, provided real-world complications are embedded into the planning architecture. A Mixed-Integer Goal Programming model is developed to enforce strict capacity thresholds and universal coverage across 832 vulnerable blocks under a simulated tsunami crisis in Callao, Peru. To maintain computational tractability, a neighborhood sparsification strategy is implemented and solved via an exact COIN-OR Branch-and-Cut engine. The empirical findings reveal that the network-based formulation converged to a tight 2.00% optimality gap in 320.31 seconds. Due to population constraints and physical accessibility barriers, the model opened 28 active distribution centers rather than the 14 initially targeted administrative regions. Crucially, ignoring real road topology introduces a severe 44.71% (166.23 kilometers) transportation underestimation bias that would lead to immediate fleet shortages during physical deployment. Ultimately, this framework bridges the gap between abstract optimization and urban friction, proving that incorporating physical street geometry and localized facility constraints is an operational prerequisite to generate dependable decision support systems for resilient disaster response.

Instituciones: Universidad del Pacífico