IIRF-EVRP

This repository contains experimental results for the paper:

"Solving Electric Vehicle Routing by Iterative Instance Refinement Framework"

The repository provides processed solution outputs and statistics for EVRP benchmark instances.

---

How to Read the Results

Each instance has a JSON file containing multiple independent runs and aggregated statistics.

Example:

solutions/E-X/E-n29-k4-s7/processed.json

---

JSON Format

Each file follows this structure:

{
  "runs": [...],
  "stats": {...},
  "global_best_*": ...
}

---

Runs

runs contains results from multiple independent executions (typically 20 runs):

{
  "run": 1,
  "best_true_total": 378.44,
  "inserted_routes": [...],
  "base_routes": [...]
}

• best_true_total: objective value of the best EV-feasible solution in this run
• inserted_routes: final EV-feasible routes (with charging stations inserted)
• base_routes: original routes before energy feasibility adjustments

---

Global Best

• global_best_true_total: best solution across all runs
• global_best_inserted_routes: corresponding EV-feasible routes

---

Statistics

"stats": {
  "n_runs": 20,
  "min": 378.44,
  "mean": 378.44,
  "std": 0.0
}

• min: best objective value across runs
• mean: average performance
• std: stability of the algorithm

---

Use

Please cite

Lin, T.-H., & Chen, Y.-p. (2026). Solving Electric Vehicle Routing by Iterative Instance Refinement Framework. In Proceedings of 2026 IEEE Congress on Evolutionary Computation (CEC 2026).