AgenticROS

AgenticROS is a framework that uses AI agents (via CrewAI) to bridge natural language task definitions with ROS robotics operations, enabling more intuitive and intelligent robot development.

Features

• 🤖 Natural language robot task definition
• 🧠 LLM-powered task decomposition and planning
• 🔄 Seamless ROS integration
• 📊 Context-aware decision making
• 🛠️ CrewAI-based cognitive architecture

Installation

1. Install ROS (tested with ROS Noetic)

# Follow ROS installation instructions for your system
# http://wiki.ros.org/noetic/Installation

2. Install Python dependencies

pip install crewai openai

3. Clone AgenticROS

cd ~/catkin_ws/src
git clone https://github.com/yourusername/agentic_ros.git
cd ..
catkin_make

4. Set up your OpenAI API key

export OPENAI_API_KEY='your-api-key-here'

Quick Start

1. Source your workspace

source ~/catkin_ws/devel/setup.bash

2. Run the example

# Terminal 1
roscore

# Terminal 2
rosrun agentic_ros test_agentic_ros.py

# Terminal 3
rosrun agentic_ros agentic_ros.py

Usage Example

from agentic_ros import AgenticRobot

# Initialize robot
robot = AgenticRobot()

# Define tasks using natural language
@agentic_task
def pick_and_place(object_name: str, location: str):
    """
    Pick up the {0} and place it at {1}.
    Be careful with the object and avoid any obstacles.
    """
    pass

# Execute tasks
robot.pick_and_place("red cube", "table A")

Architecture

AgenticROS
├── Cognitive Layer (LLM-based)
│   ├── TaskAnalyzer
│   ├── ContextManager
│   └── BehaviorGenerator
├── Bridge Layer
│   ├── ROSTranslator
│   ├── StateTracker
│   └── FeedbackProcessor
└── Execution Layer
    ├── ActionExecutor
    ├── SafetyMonitor
    └── PerformanceTracker

Contributing

We welcome contributions! Please check out our contributing guidelines and code of conduct.

License

MIT License - see LICENSE file for details

Acknowledgments

• ROS Community
• CrewAI Framework
• OpenAI

Citation

If you use AgenticROS in your research, please cite:

@misc{AgenticROS,
  author = {Kirill Balakhonov},
  title = {AgenticROS: Bridging LLMs and ROS for Intelligent Robotics},
  year = {2024},
  publisher = {GitHub},
  url = {https://github.com/NethermindEth/AgenticROS}
}

Contact

For questions and support, please open an issue or contact [kirill.balakhonov@nethermind.io]