A reinforcement learning project for solving Rubik's cubes, featuring a trained DQN agent, web API, and interactive frontend.

This project combines deep reinforcement learning with web technologies to create a complete Rubik's cube solving system:

• 🧠 RL Agent: Deep Q-Network (DQN) trained to solve Rubik's cubes
• 🚀 FastAPI Backend: REST API for cube solving
• 🎨 Vue.js Frontend: Interactive 3D cube visualization with Three.js
• 📊 Training Analytics: Comprehensive training data and performance metrics

CubeSolverRL/
├── backend/
│   ├── agent/              # DQN Agent implementation
│   ├── cube/               # Cube logic and utilities
│   ├── cube_env/           # Gymnasium environment
│   ├── models/             # Trained models (.pt files)
│   └── docs/               # Training documentation
├── frontend/               # Vue.js + Three.js frontend
│   ├── src/
│   │   └── components/     # Vue components
│   └── package.json
└── README.md

cd backend/

uv install

uvicorn api:app --reload

cd frontend/

npm install

npm run dev

The backend/docs/ directory contains detailed training analytics:

• Training curves and metrics
• Performance visualizations
• Model comparison studies
• Hyperparameter analysis

• Deep Q-Learning
• Gymnasium Documentation
• FastAPI Documentation
• Vue.js Documentation
• Three.js Documentation