Saigon300: A VRPTW Benchmark Dataset for Urban Delivery in Ho Chi Minh City

Saigon300 v1.0.1 supplements v1.0.0 with original historical evidence, corrected release documentation and reproducibility tools. The 56 instances (300 customers and one depot), eight directed OSRM matrices and archived route results are unchanged; the camera-ready paper is not edited or included. 50 m SEPARATION: The released RC geometry does not meet a global 50 m minimum: customers 18 and 161 are 22.6524193 m apart in UTM, shared by all 16 RC instances. The source files do not establish whether checks before mixing or snapping explain this. The release documents this discrepancy and does not certify the paper's global 50 m statement. RUNTIME RESOLVED: Original local log run_bih_2026-06-12.log contains two 56-instance runs. Run 2 mean is 2.4381625 seconds from logged timings, rounding to 2.44 seconds in the paper. Its six family means also round to Table 7. The previously public June 25 run records 2.327899 seconds; these are different historical runs, not conflicting measurements of one run. Both records are retained, with extracted runtime CSV and provenance JSON. Mean vehicles 20.232143, distance 1663.036339 km and travel time 2332.843452 min match the paper. ORIGINAL INPUT RESTORED: The original initial_solutions.zip was found in the author's historical experiment folder and is supplied unchanged. Its 56 initial-route sets exactly match initial_solution.routes in the archived OR-Tools outputs. This replaces an interim reconstruction prepared during audit; no reconstructed ZIP is represented as the original. June 8 initial routes differ from the later archived multi-policy BIH run and are used for comparison, not a solver warm start. VALIDATION: The new portable utility checks matrices and replays all 56 BIH and 56 OR-Tools final solutions; all pass coverage, capacity, time-window, depot-return and cost checks. All 56 initial-route sets also pass replay checks. Instructions, requirements, component licensing scope, checksums, and original runtime evidence are included. No new optimization performance experiment is claimed. REMAINING LIMITS: Dataset generation source, administrative polygons, pre/post-snap logs, original OSM extract/hash and OSRM build have not been located in the reviewed source/experiment folders. Full generation from seed 3636 and the 200 m snap bound cannot be certified. Source metadata lists verification seed 2026; archived BIH is deterministic multi-policy. Original metadata's OSM date 2026-05-26 is retained as an unverified source assertion. Initial-solution generation metadata does not certify synthetic dataset generation. RIGHTS AND SOURCE: Original CC BY 4.0 dataset and MIT code declarations are retained; upstream OSM database rights remain applicable. CC BY does not replace ODbL attribution/share-alike obligations. See LICENSING_SCOPE.txt. Attribution: https://www.openstreetmap.org/copyright . Frozen public source: https://github.com/ltlong-SGU/Saigon300-VRPTW/tree/555ac7d2eb9d2af402bdb23f9633e3af435d8ab4 . Read RELEASE_NOTES.txt and REPRODUCE.txt. Previous version DOI: 10.5281/zenodo.23949217; all-versions DOI: 10.5281/zenodo.23949216.