Decomp4OR

运筹优化模型分解对比框架 | Operations Research Decomposition Benchmarking Framework

标准化拆分集成模型为子问题流水线，公平对比求解性能 | Standardized pipeline for splitting integrated OR models into subproblems with fair performance comparison

---

描述 | Overview

运筹优化模型分解对比框架：标准化拆分集成模型为子问题流水线，公平对比求解性能。

Operations Research decomposition benchmarking framework: standardized pipeline for splitting integrated OR models into subproblems with fair performance comparison.

---

核心功能 | Features

• 模块化拆分 | Modular Decomposition：按决策域/时间/空间/资源维度拆解模型 | Split by decision domain/time/space/resource
• 标准流水线 | Standard Pipelines：集成基线、顺序求解、迭代修复、主从分解 | Integrated baseline, sequential, iterative, master-sub
• 公平对比 | Fair Comparison：统一实例、预算、随机种子 | Unified instances, budgets, random seeds
• 可复现 | Reproducible：结构化输出与配置管理 | Structured outputs and configuration management

---

快速开始 | Quick Start

1. 分析模型 | Analyze Model

识别决策变量、耦合约束、目标组成 | Identify decision variables, coupling constraints, objective components

2. 选择流水线 | Select Pipelines

• 必选 | Required：集成基线 | Integrated baseline
• 可选 | Optional：顺序/迭代/主从分解（选 1-3 条）| Sequential/iterative/master-sub (choose 1-3)

3. 执行对比 | Run Benchmark

# 配置流水线 | Configure pipelines
pipelines:
  - Integrated_Baseline
  - Sequential_A_B
  
# 运行 | Execute
python run_benchmark.py --config config.yaml

# 输出 | Outputs
results/{instance}/{pipeline}/summary.json

4. 查看结果 | View Results

python aggregate_results.py --output report.csv

---

输出格式 | Output Format

{
  "instance": "case01",
  "pipeline": "Sequential_A_B",
  "solver": {"status": "optimal", "wall": 120.5, "gap": 0.005},
  "objective": {"total": 1050.2, "components": {...}},
  "coupling": {...}
}

---

目录结构 | Directory Structure

Decomp4OR/
├── config.yaml          # 流水线配置 | Pipeline configuration
├── models/              # 模型实现 | Model implementations
│   ├── integrated.py
│   └── decomposed.py
├── pipelines/           # 流水线定义 | Pipeline definitions
├── instances/           # 实例数据 | Instance data
├── results/             # 求解结果 | Solving results
└── scripts/             # 执行脚本 | Execution scripts

---

文档 | Documentation

详见 SPEC.md 完整规范 | See SPEC.md for full specification

---

许可 | License

MIT