Dabia2013 (TDVRPTW)

Dabia2013 is the MAMUT-routing curation of the classic Duration-Minimization time-dependent VRPTW benchmark introduced by Dabia et al. 2013. It reuses the Solomon 1987 instances (25, 50, and 100 customers) and adds time dependency through the Ichoua--Gendreau--Potvin (IGP, 2003) travel-time model: piecewise-constant speeds per time zone and per arc speed profile, which yield FIFO piecewise-linear arrival-time functions. The raw JSON sources are those distributed with the open-source solver of Lera-Romero et al. 2020.

The MAMUT-routing curation replaces the historic textual formats with an explicit contract. Each instance ships the instance JSON (depot index 0, no duplicated end depot), a canonical arrival-time-function (ATF) sidecar giving one exact non-decreasing continuous piecewise-linear arrival-time function per arc of the complete graph, and a Duration BKS file. The ATFs are consolidated exactly from the raw IGP data: breakpoints are placed exactly where departure or arrival crosses a speed-zone boundary, each breakpoint is re-evaluated by the exact forward Ichoua loop, and the last zone's speed extends beyond the horizon so every ATF is total on the full time horizon.

The evaluation contract is defined by the canonical pure-Python checker (mamut_routing_lib.td.check_td_solution): route durations are computed by exact IEEE-754 double-precision composition of arrival-time and vertex ready-time functions, with no epsilon thresholds anywhere, each route dispatched at its optimal depot departure time, and the solution cost summed over routes in canonical order (sorted by first customer). BKS costs are always the checker's output.

Initial BKS were seeded from the solutions published with Lera-Romero et al. (2020) and re-validated by the checker. Several contain arrivals exactly on time-window deadlines, which fragile recomputation pipelines misreport as infeasible. Missing BKS were completed and existing ones have since been improved by KAYROS heuristic runs on Grid5000; see the family changelog.

Licensing note: MAMUT-routing-authored curation artifacts for this family are distributed under the MIT License where MAMUT-routing holds the relevant rights. The underlying benchmark definitions (Solomon 1987; Dabia et al. 2013; the Lera-Romero et al. 2020 distribution) remain third-party material and are not relicensed by this curation.

The family currently contains 168 instances over 3 sizes (56 Solomon-derived instances each at 25, 50, and 100 customers), each with a Duration BKS.

Data

Instances and best-known solutions: /benchmarks/tdvrptw/dabia2013/ on the website. Repository: https://github.com/ANR-MAMUT/MAMUT-routing/tree/main/benchmarks/TDVRPTW/Dabia2013.

Repository README

Dabia2013 TDVRPTW benchmark family (TD-IGP-Solomon)

Canonical MAMUT-routing curation of the classic Duration-Minimization TDVRPTW benchmark introduced by Dabia et al. (2013): Solomon 1987 instances (25/50/100 customers) with IGP time-dependent travel times (Ichoua et al. 2003), as distributed with the open-source solver of Lera-Romero et al. (2020).

Format

Each instance ships three artifacts per size bucket n=<customers>:

• <Name>.vrp.json — raw instance data in the MAMUT shape: depot index 0 (no duplicated end depot, contrary to historic distributions), num_customers counts customers only, mandatory coordinates, horizon, and a td block referencing the ATF sidecar with its storage-independent sha256.
• <Name>.atf.json (plain, n < 50) or <Name>.atf.json.gz (n ≥ 50) — the canonical ground truth: one arrival-time NDCPWLF per arc of the complete graph, exact consolidation of the raw IGP data. Conventions: breakpoints exactly where departure or arrival crosses a speed-zone boundary; the last zone's speed extends beyond the horizon so every ATF is total on [0, T].
• <Name>.bks.Duration.json — best known solution under the Duration objective. Costs are the authoritative output of the pure-Python canonical checker (mamut_routing_lib.td.check_td_solution): exact IEEE-754 double arithmetic, no epsilon thresholds, routes in canonical order (sorted by first customer), total summed in that order. Seeded from the published Lera-Romero et al. (2020) solutions, all re-validated by the checker (146/146 feasible; several contain arrivals exactly on time-window deadlines, which fragile recomputation pipelines misreport as infeasible).

Population pipeline: populate_td_dabia2013 v1, one-shot curation tooling maintained outside this repository (links to the migration scripts will be published with the benchmark paper); it cross-validates every consolidated ATF against direct Ichoua evaluation and reloads every written artifact through the full validation path. The generator block of every artifact records the pipeline name and version.

Provenance and licensing

Instance definitions derive from Solomon (1987) and the TD extension of Dabia et al. (2013) / Ichoua et al. (2003); the raw JSON sources are those shipped with Lera-Romero et al. (2020), https://doi.org/10.1002/net.21937. MAMUT-routing-authored curation artifacts are distributed under the repository's MIT license; this notice does not relicense the underlying third-party benchmark definitions.