Reproducibility Package for Decision-Level Counterfactual Evaluation of Selective Automation in Warehouse Order Picking

This repository contains the reproducibility package for the study “A Reproducible Decision-Level Counterfactual Evaluation Protocol for Selective Automation in Warehouse Order Picking: Cross-Configuration Failure Analysis on a Public Benchmark”. The package supports offline inspection and reproduction of the reported experiments. It includes the core Python scripts, four derived baseline event logs, state-level counterfactual results, risk-gate evaluations, a data dictionary, provenance information, and a SHA-256 manifest. The study uses two parameter configurations from the KIT Manual Warehouse Order Picking public benchmark, covering 200 complete shifts and 164,052 orders. The benchmark is a public synthetic instance and does not represent real warehouse management system event logs. The complete upstream KIT archive is not redistributed in this repository. It can be obtained from the published source under its applicable licence: https://doi.org/10.35097/mwsv59v8sk9sqaan The README.md and REPRODUCTION.md files provide the environment requirements and reproduction commands. The fixed project audit returned a NO_GO decision. This package does not claim real-world warehouse performance, causal effects, deployment readiness, or guaranteed operational safety.