robot_agent

Chinese

Introduction

This project uses LangChain / LangGraph to connect an LLM with robots under ROS: you send natural-language commands in the terminal; the model reasons and calls low-level tools (ROS queries, flight commands, etc.) to drive the drone.

Installation

git clone https://github.com/mono0826/drone_ws.git
cd drone_ws
catkin_make
source devel/setup.bash

robot_agent

git clone https://github.com/mono0826/robot_agent.git
cd robot_agent
pip install -e .
pip install -r requirements.txt

Copy .env.example to .env and set your API keys.

Usage

roslaunch simulator drone_sim_1uav.launch
roslaunch px4ctrl run_ctrl.launch
roslaunch px4ctrl terminal_ctrl.launch

In another terminal:

cd robot_agent
python examples/drone.py

Demo

Roadmap

• Chat history: Persist multi-turn dialogue or combine with summarization for long tasks and review.
• Vision-language navigation: Use camera / depth (and related topics) with a VLM for scene understanding and high-level navigation, wired into the existing control stack.