Data and code for "When are delivery drones worth retaining? Multimodal last-mile planning across terrain and access conditions"

This record contains the research inputs, model code, saved route-search outputs, customer-level access screens, hub-layout and access-envelope sensitivities, small-instance exact-check outputs, and separately labelled route-level stress diagnostics supporting the manuscript “When are delivery drones worth retaining? Multimodal last-mile planning across terrain and access conditions.” The study uses four contrasting road-network cases and twelve spatially resampled sets of 120 synthetic customers. Customer locations are sampled from mapped network nodes; they are not observed customer addresses or carrier orders. Operating costs, service rules, energy rates, and emissions are model scenarios, not field-measured delivery outcomes. The large-instance heuristic reports best-found plans, not certified global optima. The release is frozen to the evidence lineage used by the manuscript. Later demand-release and route experiments in the working research archive are not included as evidence for the manuscript. The archive includes OpenStreetMap-derived road-network inputs, Taiwan Ministry of the Interior 20 m DTM-derived processed road-edge fields, Amazon-derived planned service-duration values, original code, synthetic instances, and generated results. It does not redistribute raw DTM rasters or the original Amazon dataset. See README.md and LICENSES_AND_ATTRIBUTION.md inside the archive for provenance, reuse conditions, and reproduction notes.