Spot Mini Mini - PyBullet Simulation Setup Guide

This guide walks through setting up and running the OpenQuadruped/spot_mini_mini project on Ubuntu 22.04 using Python 3.10 (ROS not required).

---

Prerequisites

Ensure the following are installed:

sudo apt update && sudo apt install -y python3-pip python3-venv git xvfb x11-utils libgl1-mesa-glx libegl1-mesa

---

Create Python Virtual Environment

python3 -m venv ~/spot_env
source ~/spot_env/bin/activate

Install required Python packages:

pip install numpy gym pybullet matplotlib
pip install pyyaml typeguard

---

Clone the Repository

cd ~
git clone https://github.com/OpenQuadruped/spot_mini_mini.git

---

Test the Environment

Run the GUI simulator to check if the environment launches:

cd ~/spot_mini_mini/spot_bullet/src
python3 env_tester.py

You should see a GUI with the Spot robot standing. This confirms your environment is working.

---

Run Pretrained Walking Model

The repo contains pretrained models.

Run an evaluation agent (example):

cd ~/spot_mini_mini/spot_bullet/src
python3 spot_ars_eval.py -a 149

To view other models:

ls ~/spot_mini_mini/models/

Use the filename number with -a <num> (e.g. -a 100, -a 149, etc.)

---

Fixes Made

• Edited Bezier.py to fix NumPy error:

  Replaced:

  np.math.factorial(...)

  With:

  import math
  math.factorial(...)

• Installed missing packages:

  pip install pyyaml typeguard

---

Training Your Own Gait Model

(Coming soon or optional extension by contributors)

To enable gait experimentation:

• Edit gait logic in: spotmicro/GaitGenerator/Bezier.py
• Use env_tester.py for real-time visual testing
• Train with spot_train.py (after adding CLI gait type support)

---

Debug Tips

• If GUI launches but robot doesn't walk, check:
  • You’re loading a working model number
  • Bezier config in Bezier.py isn’t broken
• If training throws error: double-check NumPy, math, or missing dependencies
• If GUI crashes: install missing OpenGL libs

---

Contribution Guide

• Fork the repo and create a new branch
• Test your gait in env_tester.py
• Log your results and submit PR
• Add model weights to models/ with matching ID

---

Summary of Key Commands

# Environment setup
python3 -m venv ~/spot_env
source ~/spot_env/bin/activate
pip install numpy gym pybullet matplotlib pyyaml typeguard

# Clone repo
git clone https://github.com/OpenQuadruped/spot_mini_mini.git

# Test environment
cd ~/spot_mini_mini/spot_bullet/src
python3 env_tester.py

# Evaluate pretrained walking model
python3 spot_ars_eval.py -a 149

# Train new model (advanced)
python3 spot_train.py --gait_type trot  # (after adding gait_type support)

---

Maintainer Notes

If you're editing:

• Bezier.py: Handles trajectory planning
• ars.py: Reinforcement learning agent
• spot_train.py / spot_ars_eval.py: Entry points

Ensure to update this README and test thoroughly. [

Your browser does not support the video tag. ](https://github.com/cyan-ide7/ros_quadrupedal/blob/main/Screencast%20from%2007-08-2025%2001_01_26%20AM%20(online-video-cutter.com).mp4?raw=true )