Welcome to the Hunting Game, an interactive single-player game where you play as the prey or predator against trained AI agents!

This project showcases the power of Multi-Agent Reinforcement Learning (MARL) using a custom environment and agents trained with Independent Proximal Policy Optimization (IPPO), all wrapped in a web-based interface.

Hunting Game animation

🎮 Gameplay Features

  • 🧠 Smart AI agents trained with JAXAgents, JAX, Flax, and Optax
  • 🎭 Play as prey or predator against AI with unique strategies
  • 🕹️ Cross-platform controls: keyboard or on-screen joystick
  • 🌐 Browser-based — no installation required
  • 🔄 Replayable with different roles and strategies
  • 📹 Game recording for agent fine-tuning (coming soon)

🧠 Tech Stack

Component Technology
RL Agents JAXAgents, JAX, Flax, Optax
Algorithm Improved PPO (IPPO)
Environment Custom prey-predator grid setup
Frontend HTML, CSS, JavaScript
Mobile Input On-screen joystick
Backend API FastAPI
Containerization Docker
Hosting GitHub Pages (frontend), Hetzner Cloud (backend)

💡 Roadmap

  • 🎮 Add human vs. human multiplayer mode
  • 📈 Implement curriculum learning for progressive difficulty
  • 📊 Create analytics dashboard (reward curves, agent heatmaps)
  • 🎨 Polish UI/UX with animations and tutorials

Made with passion for gaming and AI

Trained using JAXAgents based on JAX & Flax, and deployed via FastAPI.