🏃 Sports Hub

🔗 Check out the website for live demo's here

Sports Hub is a simple collection of clean, high-performance Locomotion MJX environments built for training reinforcement learning agents using Brax + JAX.

🚀 Plug in, pick an environment, and watch your agent go.

📦 Features

• ⚡ Ultra-fast training with JAX + Brax
• 🎮 Easy-to-run, modular MJX environments
• 🔁 Reproducible with RNG seed control
• 🎬 Sample trained agents to visualize behavior
• 🛠 Built to extend: plug in your own logic, bodies, and control

🌍 Supported Environments

Environment	Description	Status	Docs 🧍	Docs 🐜
sprint	Agent sprinting across a race track	🟢	📄	📄
longjump	Agent attempting the long jump field event	🟢	📄	📄
gaps	Agent jumps across a platform of gaps	🟢	📄	➖
hurdles	Agent sprints and jumps over hurdles	🟢	📄	➖
fetch	Agent fetches a ball and brings it to target	🟢	📄	📄
ball dribble	Agent dribbles a soccer ball to the finish	🟣✨	➖	➖

Legend: 🟢 Fully supported  🟣✨ Coming Soon  🧍 Humanoid Variant  🐜 Quadruped Variant

⚡ Quickstart

Clone the repo and set up your environment:

git clone https://github.com/i1Cps/sports-hub.git
cd sports-hub 

python3 -m venv env
source env/bin/activate

🧠 Install JAX

Choose one of the following based on your system:

• If you have a GPU (faster):

  pip install jax[cuda12]
  

• If you don't have a GPU (CPU only):

  pip install jax
  

📦 Install the rest of the dependencies

pip install -r requirements.txt

Install VLC

sudo apt install vlc

🚀 Run an example agent

python3 -m sports_hub.inference --env=sprint_humanoid --rng=3 --num_agents=5

🧪 Swap --env= with any supported task
🧪 Use --num_agents to control how many agents are spawned (⚠️running more than one agent can destabilize the simulation)
🧪 Use --rng for different seeds

📦 Dependencies

🙏 Credits

Developed by Theo Moore-Calters

Special thanks to

📜 License

This project is licensed under the MIT License.
See the LICENSE file for details.