alirezaaminzadeh/Deep-RL-for-Dynamic-Job-Shop-Scheduling-GNN-Transformer-Multi-Agent-System

🤖 RL-Scheduler: Deep RL for Job Shop Scheduling

A state-of-the-art reinforcement learning system combining Graph Neural Networks (GNN), Transformer architectures, and Multi-Agent RL to solve the Dynamic Job Shop Scheduling Problem (DJSP).

🎯 Overview

This interactive demo showcases how deep learning can solve complex operations research problems. The system uses:

• —Graph Neural Networks to encode the job shop topology
• —Transformer-based policies for attention-based scheduling decisions
• —Multi-Agent RL for decentralized coordination
• —Hybrid optimization combining neural networks with classical OR solvers

🚀 Features

• —✅ Interactive Gradio interface for real-time scheduling
• —✅ Multiple algorithms: RL Agent, SPT, FIFO
• —✅ Gantt chart visualization
• —✅ Configurable problem size (jobs & machines)
• —✅ Performance metrics comparison

📊 Performance Benchmarks

🛠️ Technology Stack

• —Deep Learning: PyTorch, PyTorch Geometric
• —RL: Stable-Baselines3, Gymnasium
• —OR Tools: Google OR-Tools
• —UI: Gradio, Plotly
• —Deployment: Docker, FastAPI, ONNX Runtime

💡 How to Use

1. 1.Configure Problem: Set number of jobs and machines
2. 2.Select Algorithm: Choose between RL agent or classical heuristics
3. 3.Solve: Click the solve button to generate optimal schedule
4. 4.Visualize: View Gantt chart and performance metrics

📚 About

This project demonstrates both academic rigor and commercial viability by applying cutting-edge deep learning techniques to classical operations research problems.

Author: Alireza Aminzadeh Email: alireza.aminzadeh@hotmail.com GitHub: @syeedalireza License: MIT

🔗 Links

• —Full Documentation
• —Paper (Coming Soon)
• —API Documentation

📖 Citation

@software{aminzadeh2026rlscheduler,
  author = {Aminzadeh, Alireza},
  title = {RL-Scheduler: Deep Reinforcement Learning for Dynamic Job Shop Scheduling},
  year = {2026},
  url = {https://huggingface.co/spaces/syeedalireza/Deep-RL-for-Dynamic-Job-Shop-Scheduling-GNN-Transformer-Multi-Agent-System}
}