This project provides a set of helper scripts and a low-cost ($300) robot arm (SO-100) designed to make robotics accessible for experimentation and application discovery. It targets hobbyists, researchers, and developers looking to build and deploy custom robotic tasks, leveraging foundational models and enabling multi-model task execution.

How It Works

The system builds upon the LeRobot framework, integrating helper scripts for enhanced functionality. Its core innovation lies in enabling the chaining of multiple AI models (e.g., ACT models) to solve complex robotics tasks sequentially. This modular approach allows for specialized models to handle different sub-tasks, such as object grasping and precise insertion, improving overall task performance and flexibility.

Quick Start & Requirements

• Installation: Clone LeRobot (git clone https://github.com/huggingface/lerobot.git), install (cd lerobot; pip install -e .), then clone SimpleAutomation (git clone https://github.com/1g0rrr/SimpleAutomation.git).
• Prerequisites: Python, LeRobot framework. Robot configuration (core/configs/robot/so100.yaml) is required for robot-specific ports.
• Resources: Training models can be done in Google Colab (approx. 2.5 hours, $1.5 with A100 GPU).
• Links: LeRobot: https://github.com/huggingface/lerobot, Training Colab: https://colab.research.google.com/github/1g0rrr/SimpleAutomation/blob/main/colab/SimpleAutomationTrainModel.ipynb

Highlighted Details

• Supports chaining multiple AI models for complex task execution (e.g., lamp testing demo).
• Includes teleoperation features for enhanced data recording between episodes.
• Offers an LLM agent for model interaction.
• Provides a low-cost SO-100 robot arm option.

Maintenance & Community

• Active community via Discord: https://discord.gg/NFsqq4CVhs

Licensing & Compatibility

• The README does not explicitly state a license. Compatibility for commercial use or closed-source linking is undetermined.

Limitations & Caveats

The project is in an early stage, with the README noting that pick-and-place tasks can be challenging for models, potentially requiring re-grasping logic. Initial positions must be carefully configured to prevent unexpected robot movements.