MOHAF: Multi-Objective Hierarchical Auction Framework

Overview

This repository provides the implementation and supporting materials for the paper:

"MOHAF: A Multi-Objective Hierarchical Auction Framework for Scalable and Fair Resource Allocation in IoT Ecosystems"

MOHAF is a distributed resource allocation mechanism for heterogeneous IoT environments.
It combines hierarchical clustering, submodular optimization, and dynamic pricing to achieve:

• Efficient allocation of resources in large-scale IoT ecosystems
• Joint optimization of cost, energy efficiency, quality of service (QoS), and fairness
• Theoretical guarantees with a (1 - 1/e) approximation ratio
• Perfect fairness across participants (Jain’s Index = 1.000 in experiments)

We provide code, experimental setup, and references to enable reproducibility and extension of this work.

---

Repository Contents

MOHAF-Resource-Allocation
├─ .gitignore
├─ LICENSE
├─ MOHAF.ipynb
├─ README.md
├─ docs
│  ├─ index.md
│  ├─ mohaf_paper.pdf
│  └─ usage.md
├─ examples
│  ├─ analyze_results.ipynb
│  └─ run_experiments.py
├─ mohaf
│  ├─ __init__.py
│  ├─ auction_mechanisms.py
│  ├─ core.py
│  ├─ scenarios.py
│  └─ utils.py
├─ pyproject.toml
├─ requirements.txt
└─ tests
   ├─ __init__.py
   ├─ test_auction_mechanisms.py
   └─ test_core.py

---

Installation

You can install MOHAF directly from PyPI:

pip install mohaf

Usage

After installation, you can import MOHAF in your Python scripts:

from mohaf.auction_mechanisms import MOHAFAuction
from mohaf.scenarios import generate_synthetic_scenario

# 1. Initialize the auction mechanism
mohaf_auction = MOHAFAuction(alpha=0.3, beta=0.3, gamma=0.2, delta=0.2)

# 2. Generate a synthetic scenario
resources, requests = generate_synthetic_scenario(n_resources=50, n_requests=30)

# 3. Run the auction
results = mohaf_auction.run_auction(resources, requests)
metrics = mohaf_auction.calculate_metrics(results)

print("Auction completed!")
print(f"  Allocation Efficiency: {metrics['allocation_efficiency']:.3f}")
print(f"  Revenue: ${metrics['revenue']:.2f}")

For more detailed examples and to reproduce the experiments from the paper, please see the examples/ directory and the MOHAF.ipynb notebook in the repository.

---

Command Line Interface (CLI)

You can also run MOHAF from the terminal without writing code once installed from PyPI.

Basic usage

# Default MOHAF mechanism with a balanced synthetic scenario
mohaf

# Specify number of resources/requests and scenario type
mohaf --n-resources 100 --n-requests 60 --scenario-type high_demand

# Choose a different baseline mechanism
mohaf --mechanism greedy

Available mechanisms

• mohaf (default)
• first_price
• vickrey
• hungarian
• greedy
• random

MOHAF weights

You can tune the MOHAF multi-objective weights and learning rate:

mohaf --alpha 0.3 --beta 0.3 --gamma 0.2 --delta 0.2 --learning-rate 0.01

Reproducibility and loading scenarios from file

• Seed: --seed 42 to make random generation deterministic.
• Load from JSON: Provide a file with resources and requests arrays matching the mohaf.core.Resource and mohaf.core.Request fields. Example structure:

{
  "resources": [
    {
      "id": "resource_0",
      "type": "compute",
      "capacity": 50,
      "cost_per_unit": 0.8,
      "location": [0, 0],
      "availability": 0.9,
      "reliability": 0.85,
      "energy_efficiency": 0.75,
      "owner_id": "owner_0"
    }
  ],
  "requests": [
    {
      "id": "request_0",
      "requester_id": "device_0",
      "resource_type": "compute",
      "amount": 20,
      "max_price": 30,
      "deadline": 1730000000.0,
      "priority": 5,
      "location": [10, 10],
      "qos_requirements": {"min_reliability": 0.6, "max_latency": 50}
    }
  ]
}

Run with:

mohaf --input scenario.json

JSON output

Return results and metrics as JSON (for scripting/automation):

mohaf --json --pretty

This prints

{
  "result": { "allocations": [...], "execution_time": 0.123, ... },
  "metrics": { "allocation_efficiency": 0.87, "revenue": 123.45, ... }
}

Saving results to files

• Allocations to CSV: --allocations-csv allocations.csv
• Metrics to JSON: --metrics-json metrics.json

Quiet mode

Suppress internal prints from mechanisms:

mohaf --quiet

Full help

mohaf --help

Output includes all flags:

• --mechanism {mohaf,first_price,vickrey,hungarian,greedy,random}
• --n-resources INT and --n-requests INT
• --scenario-type {balanced,high_demand,low_resource}
• --alpha --beta --gamma --delta --learning-rate (MOHAF only)
• --json and --pretty

---

Dataset

Experiments rely on the publicly available Google Cluster Data: https://github.com/google/cluster-data

---

License

This project is licensed under the MIT License.

---

Citation

If you use this repository, please cite the associated paper:

@misc{Agrawal2025,
  author       = {Agrawal, Kushagra and Goktas, Polat and Bandopadhyay, Anjan, and Ghosh, Debolina and Jena, Junali Jasmine and Gourisaria, Mahendra Kumar},
  title        = {MOHAF: A Multi-Objective Hierarchical Auction Framework for Scalable and Fair Resource Allocation in IoT Ecosystems},
  year         = {2025},
  eprint       = {2508.14830},
  archivePrefix= {arXiv},
  primaryClass = {cs.DC},
  doi          = {10.48550/arXiv.2508.14830},
  url          = {https://arxiv.org/abs/2508.14830}
}

---

📌 Repository link: https://github.com/afrilab/MOHAF-Resource-Allocation/tree/main