ros-prompt

Natural language robot control for ROS 2 using LLMs (LangChain).

ROS Prompt Demo: Autonomously switch between SLAM and Navigation mode (click for YouTube)

Features

• Understands and plans robot actions from natural language
• Automatic capability discovery (topics, actions, services)
• Intent detection and behavior tree generation
• Modular and extensible ROS 2 package

Getting Started

1. Run export ROS_PROMPT_OPENAI_KEY="sk-..." with your own API key.
2. pip install py_trees langchain langchain_openai jsonschema
3. Clone ros-prompt into your ROS workspace
4. colcon build --packages-select ros_prompt
5. source install/setup.bash
6. Start your launch files (ie. ros2 launch turtlebot3_gazebo turtlebot3_world.launch.py)
7. ros2 launch ros_prompt ros_prompt.launch.py
8. Publish your messages to /user_input: ros2 topic pub /user_input std_msgs/String "data: 'start slam, explore, save map, swith to nav with map, set initial pose, wait for 12 seconds, navigate to position (0.6, 1.5)'" -1

License

MIT