A Multi-Warehouse Multi-Tour Drone Routing Model for Post-Disaster Relief Distribution:Metaheuristics and a Case Study of İstanbul
Authors:
- Ozan Baran DemirçiviODOzan Baran Demirçivi
- Onat DüzgünODOnat Düzgün
- Arya Sevgen MisicAMArya Sevgen Misic
- Selin OzpeynirciSOSelin Ozpeynirci
Preprint
- 27 October 2026
- preprint
- Published by Elsevier in SSRN Electronic Journal
Abstract
Post-disaster humanitarian logistics operations frequently face severe gridlocks in metropolitan environments due to extensive road network damage and debris accumulation. Under such operational paralysis, Unmanned Aerial Vehicles (UAVs) provide a resilient alternative for last-mile emergency relief distribution. This study addresses a synchronized, multi-warehouse, multi-UAV, multi-tour vehicle routing problem tailored to urban post-disaster logistics. This involves capacity-constrained drones performing multiple consecutive delivery tours subject to mandatory warehouse battery-exchange and reloading turnaround times. To prioritize critical emergency demands, we formulate a Mixed-Integer Linear Programming (MILP) model that minimizes the total priority-weighted arrival time, explicitly incorporating urgency scores derived from fault-line proximity. Given the combinatorial NP-hardness of the underlying problem, we develop a multi-stage metaheuristic architecture consisting of a constraint-aware Greedy Randomized Adaptive Search Procedure (GRASP) to generate executable seed topologies. This approach is complemented with two advanced search engines: a population-based Lamarckian Hybrid Genetic Algorithm (HGA) embedding memetic local-search overwrites, and a single-trajectory Dynamic Hybrid Neighborhood Search (HNS) featuring cataclysmic disruption mechanics to escape deep local optima. The proposed methodologies are rigorously calibrated via Taguchi Design of Experiments and evaluated on benchmark instances alongside a realistic large-scale earthquake case study of İstanbul comprising 6 depots and 171 assembly areas. Computational experiments demonstrate that both custom heuristics match CPLEX global optimal baselines on small instances, and, on medium and large problem instances, systematically outperform the state-of-the-art commercial solver Hexaly with HNS securing 68.7% of the best-known solutions. Furthermore, managerial analyses reveal critical operational thresholds in drone fleet sizing and underscore the vital role of dynamic inter-depot fleet relocation in post-disaster response.Keywords
This publication has 35 references indexed in Scilit:
- Multiple traveling salesman problem with drones: Mathematical model and heuristic approachComputers & Industrial Engineering, 2019
- Same‐day delivery with heterogeneous fleets of drones and vehiclesNetworks, 2018
- Drone delivery from trucks: Drone scheduling for given truck routesNetworks, 2018
- The vehicle routing problem with drones: Extended models and connectionsNetworks, 2017
- The vehicle routing problem with drones: several worst-case resultsOptimization Letters, 2016
- Efficient route planning for an unmanned air vehicle deployed on a moving carrierSoft Computing, 2015
- Post-earthquake response by small UAV helicoptersNatural Hazards, 2015
- Record Breaking Optimization Results Using the Ruin and Recreate PrincipleJournal of Computational Physics, 2000
- Greedy Randomized Adaptive Search ProceduresJournal of Global Optimization, 1995
- Various optimizers for single‐stage productionNaval Research Logistics Quarterly, 1956