🚚 3D Box Packing + GA Route Optimization (Dissertation Project)

This repository contains the full implementation of my dissertation work on optimizing vehicle loading and routing using:

• Genetic Algorithms (GA) for route ordering
• 3D Box/Bin Packing (3D-BPP)
• Iterative Isolation / VLR Repair methods
• Automated replicate experiments
• Timestamped experiment pipelines
• Full reproducibility of results

The system reads real-world merged & normalized datasets, runs multiple GA replicates, applies spatial repair, summarizes results, and generates publication-ready figures.

---

📁 Project Structure

Dataset Generation/
│
├── input_dataset/
│     └── XML100_1111_01_merged_with_boxes_norm.json
│
├── experiments/
│     └── run_replicates.ps1
│
├── scripts/
│     ├── summarize_replicates_fixed.py
│     ├── extract_isolations.py
│     └── make_figures_fixed.py
│
├── dataset_generation/
├── results/
│     └── runs/
│
└── run_all_experiments.ps1

---

🎯 Project Purpose

This project evaluates a hybrid optimisation pipeline combining:

• GA-driven route/order optimisation
• Realistic 3D box/bulk packing
• Automatic infeasibility detection & repair
• Multi-replicate GA runs
• Automated summary & figure generation

---

⚙️ Setup Instructions

1️⃣ Clone the repository

git clone <your-repo-url>
cd "Dataset Generation"

2️⃣ Create and activate a virtual environment

python -m venv .venv
.\.venv\Scripts\Activate.ps1

3️⃣ Install dependencies

If using requirements:

pip install -r requirements.txt

Otherwise:

pip install numpy pandas matplotlib

4️⃣ Add your dataset file

Place your dataset here:

input_dataset/XML100_1111_01_merged_with_boxes_norm.json

---

🚀 Running the Project

✅ Option A: Full Automated Pipeline (Recommended)

Runs the full workflow and saves results in timestamped folders:

.\run_all_experiments.ps1

This script automatically:

• Loads dataset
• Runs GA replicates
• Saves everything under results/runs/<timestamp>/
• Generates: summary CSV, isolation CSV, figures, logs

---

✅ Option B: Run GA Replicates Manually

.\experiments\run_replicates.ps1 `
    -normfile ".\input_dataset\XML100_1111_01_merged_with_boxes_norm.json" `
    -replicates 20 `
    -seed_start 2000

Outputs go to:

experiments_output/

---

📊 Output Files Explained

✔ replicates_summary.csv

Per-seed GA performance (best score, duration, unpacked, infeasible, etc.)

✔ figures/

Automatically generated plots:

• best_score vs seed
• duration vs seed
• unpacked distribution
• infeasible histogram

✔ isolation_summary.csv

Shows spatial-repair / isolation behaviour.

✔ summary_stats.txt

Contains timestamp, params, and output paths.

✔ experiments_output/

Raw per-seed JSON outputs.

---

📌 Recommended .gitignore

# Python
__pycache__/
*.pyc

# Virtual environment
.venv/

# Output folders
experiments_output/
results/
output/

# Logs
*.log

# Large datasets
input_dataset/*.json
input_dataset/*.csv

# Backup files
*.bak

---

🧠 Key Concepts Implemented

✔ Genetic Algorithm (GA)

Optimises route/stop ordering; multiple seeds measure stability.

✔ 3D Packing + Repair

Uses rotation rules, feasibility checks, and VLR/Isolation repair.

✔ Replicate-Based Robustness

Shows variability, convergence stability, and runtime consistency.

✔ Automatic Analysis

Generates figures and summaries for dissertation use.

---

🧪 Reproducibility

To reproduce all experiments:

.\run_all_experiments.ps1

Outputs will appear under:

results/runs/<timestamp>/

---