JOBASRPRDTW benchmark: instances, non-dominated fronts, hypervolume curves and analysis scripts (60 dataset–instance blocks, 8 methods, 30 seeds)

Benchmark instances and full campaign results for JOBASRPRDTW, a biobjective integrated order picking problem that couples order batching, picker–cart assignment and sequencing, picker routing, reusable roll-cage carts, dock time windows and a workload-balance objective. The archive contains 120 instance and evaluation fixtures (6 datasets × 10 instance seeds), the final non-dominated fronts of 14,400 runs (8 methods × 30 algorithmic seeds × 60 dataset–instance blocks), per-generation hypervolume curves, decoded logistics of the selected representative solutions, per-block configurations and quality gates, the geometric post-processing of the 12 high-resolution blocks, the record of the ex ante calibration, and the scripts that verify the artifacts and regenerate the article's data-derived tables and figures. All solutions were evaluated with a deterministic decoder on a CUDA GPU backend, so the evaluation is identical across the eight compared methods. Every file is covered by MANIFEST.sha256; running scripts/verify_artifacts.py re-checks the checksums and the structural invariants of the campaign. Dual license: code under MIT, data and figures under CC-BY-4.0. See LICENSE. These data support the article “Adaptive selection pressure for GPU-accelerated decomposition-based evolutionary optimization of the biobjective integrated order picking problem with dock windows and reusable resources”, submitted to Advanced Engineering Informatics. The DOI of the article is added to the metadata once it is assigned.