ROSGPT: ROS Guide Powered by Transformers

ROSGPT is a framework that connects Large Language Models (LLMs) with Robot Operating System (ROS) to enable natural language control of robotic systems.

Project Structure

The project consists of several components:

1. ROSGPT Framework: Core framework that provides LLM integration with ROS tools and capabilities
2. TurtleSim Agent: A demonstration agent that controls a turtle in the ROS TurtleSim environment
3. PincherX-100 Agent: An agent that controls a PincherX-100 robotic arm

Requirements

• Docker
• Docker Compose
• X11 for GUI (not needed for headless mode)
• Groq API key

Getting Started

1. Clone this repository:

   git clone https://github.com/uefroboticsclub/rosgpt.git
   cd rosgpt
   

2. Copy the environment example file and set your Groq API key:

   cp .env.example .env
   # Edit .env to add your Groq API key and other settings
   

3. Choose which agent to run:

Running the TurtleSim Agent

./demo.sh

This will set up the Docker environment for the TurtleSim. Once inside the container, build and start the turtle agent:

start

You can enable streaming mode by passing the streaming argument:

start streaming:=true

Running the PincherX-100 Agent

./pincherx_demo.sh

This will start the Docker container for the PincherX-100 and connect you to the agent.

Environment Variables

You can customize the agent behavior by setting these variables in your .env file:

• GROQ_API_KEY: Your API key for Groq
• GROQ_MODEL: The model to use (default: "mixtral-8x7b-32768")
• LLM_TEMPERATURE: Temperature for model sampling (default: 0.7)
• LLM_MAX_TOKENS: Maximum tokens for model response (default: 8192)

Agent Command Reference

Both agents support the following basic commands:

• help: Display help and usage information
• examples: Show and select from example commands
• clear: Clear the chat history
• exit: Exit the agent

TurtleSim Agent Commands

The TurtleSim agent can understand natural language instructions to:

• Draw shapes
• Move the turtle in specific directions
• Change colors
• Teleport the turtle to specific coordinates

PincherX-100 Agent Commands

The PincherX-100 agent has these additional commands:

• home: Move the arm to home position
• sleep: Move the arm to sleep position

The agent can understand natural language instructions to:

• Move to specific joint positions
• Move the end-effector to specific poses
• Open and close the gripper
• Pick and place objects
• Execute complex trajectories

Docker Environments

The project uses two separate Docker environments:

1. TurtleSim: Uses ROS1 (Noetic) for the TurtleSim agent
2. PincherX-100: Uses ROS2 (Humble) for the PincherX-100 agent

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

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