SO-101 Playground

An experimental workspace for robot learning research with the SO-101 robot arm.

Overview

🚧 Work in Progress

This repository contains tools and experiments for:

• 🤖 Imitation Learning - Record and replay demonstrations using LeRobot
• 🎮 Deep Reinforcement Learning - Train policies with RL algorithms
• 🧠 Vision-Language-Action Models - Explore VLA-based control
• 🔬 Ablation Studies - Systematic experiments on task design and architecture

Hardware Setup

Component	Port/Device
Follower Arm	/dev/ttyACM0
Leader Arm	/dev/ttyACM1
Robot Camera	/dev/video2

Installation

# Clone the repository
git clone git@github.com:vvrs/so101-playground.git
cd so101-playground

# Create virtual environment
python -m venv .venv
source .venv/bin/activate

# Install dependencies (includes LeRobot from source)
pip install -e .

# Grant serial port access
sudo usermod -aG dialout $USER
# Log out and back in for group changes to take effect

Quick Start

1. Calibrate the Arms

# Calibrate follower arm
lerobot-calibrate \
    --robot.type=so101_follower \
    --robot.port=/dev/ttyACM0 \
    --robot.id=my_follower_arm

# Calibrate leader arm
lerobot-calibrate \
    --teleop.type=so101_leader \
    --teleop.port=/dev/ttyACM1 \
    --teleop.id=my_leader_arm

2. Teleoperate

lerobot-teleoperate \
    --robot.type=so101_follower \
    --robot.port=/dev/ttyACM0 \
    --robot.id=my_follower_arm \
    --teleop.type=so101_leader \
    --teleop.port=/dev/ttyACM1 \
    --teleop.id=my_leader_arm

3. Record Demonstrations

python record.py --episodes 10 --task "Pick and place object"

4. View Camera

python view_camera.py --device 2

Project Structure

so101-playground/
├── record.py          # Recording script for demonstrations
├── view_camera.py     # Camera viewer utility
├── CHEATSHEET.md      # Quick reference for common commands
├── pyproject.toml     # Project configuration
└── ablations/         # Experimental studies

License

MIT