The Geometric Directed Chinese Postman Problem under Angle-Monotone Turn Costs

Description

This dataset provides the complete reproducibility package for the research article titled "The Geometric Directed Chinese Postman Problem under Angle-Monotone Turn Costs", submitted to Transportmetrica A: Transport Science. It contains all Python source code, test suites, synthetic instance generators, and raw experimental output logs required to fully reproduce every table, figure, and statistical result presented in the manuscript. The package includes a modular implementation of the two-stage approximation algorithm for the Geometric Directed Chinese Postman Problem (Geo-DCPP), along with scripts for scaling studies, robustness analysis under axiom violations, and extensive verification of the Separability Conjecture through over 500 perturbation tests across grids, radial networks, and one-way systems. Additional tools are provided for automatically generating publication-ready tables and visualisations that match those in the paper. Users can reproduce all results by running a single main script. The dataset also includes a fully functional OSMnx-based pipeline for extracting and processing real-world road networks from OpenStreetMap, enabling future application of the method to practical urban operations such as street sweeping, snow removal, refuse collection, and municipal fleet routing. This release ensures complete transparency and reproducibility of the reported approximation ratios (consistently below 1.013, with a mean of 1.005), the linear scaling of the optimality gap with the number of merges, and the strong robustness of the algorithm even when the angle triangle inequality is violated. The code is developed in Python 3.10+ following scientific computing best practices and is released under the CC BY 4.0 licence.

Steps to reproduce

Steps to Reproduce the Results Environment Setup Python 3.10 or higher is required. Install dependencies: pip install -r requirements.txt Reproduce All Tables and Figures Run the main script: python py/empirical_study.pyThis script automatically generates: Table 1 (Scaling study) Table 2 (Robustness to axiom violations) Figure 4 (Ratio vs |A| and Gap vs Merges) All separability and robustness results Individual Studies (Optional) You can also run the studies separately: python py/scaling_study.py → Scaling results (Table 1) python py/robustness_study.py → Robustness results (Table 2) python py/separability_study.py → Separability conjecture tests python py/adversarial_separability.py → Extreme A3 violation tests Reproduce Statistical Tests python py/gap_slope_study.py → Linear regression on gap vs merges (Figure 4 right panel) Run Unit Tests (Recommended) python -m pytest tests/ -v Output Location All generated tables, figures, and raw results are saved in the results/ folder.

Institutions

• Texas Tech University

  Texas, Lubbock

Categories

Licence