caatingarobotics

ROS 2 workspace for precision agriculture robotics, combining simulation, navigation, computer vision inference, and field traceability analytics.

Project Pitch

This project addresses a practical agriculture challenge: inspect large crop areas, detect infestation signals, and support targeted intervention with fewer inputs.

The solution integrates:

• a ROS 2 simulation stack for robot operation and mission flow,
• a YOLO-based vision pipeline for on-board detection,
• analytics and traceability outputs to support agronomic decisions.

Architecture Overview

The workspace is organized in two main ROS 2 packages:

• agro_robot_sim: robot model, Gazebo worlds, SLAM/navigation launch flows, mission scripts.
• caatinga_vision: camera source, YOLO inference, infestation analytics, photo capture pipeline.

High-level data flow:

1. Sensor/camera data is published.
2. Vision nodes run inference and publish detections.
3. Analytics nodes aggregate detections and generate status/recommendations.
4. Navigation and mission layers consume maps/localization for route execution.

Tech Stack

• ROS 2 Humble
• Gazebo + Nav2 + SLAM Toolbox
• YOLO (Ultralytics), OpenCV, cv_bridge
• Python (rclpy) and CMake (ament_cmake)
• PyQt5 operational panel tooling

Repository Structure

.
├── datasets/
│   └── agro_v1/
│       └── data.yaml
├── src/
│   ├── agro_robot_sim/
│   │   ├── launch/
│   │   ├── config/
│   │   ├── maps/
│   │   ├── urdf/
│   │   └── worlds/
│   └── caatinga_vision/
│       ├── caatinga_vision/
│       ├── launch/
│       ├── config/
│       └── models/
└── README.md

Quick Start

1. Prerequisites

• Ubuntu 22.04
• ROS 2 Humble
• colcon
• rosdep

2. Build

cd /home/joaodemoura/agro_robot_ws
source /opt/ros/humble/setup.bash
rosdep update
rosdep install --from-paths src --ignore-src -r -y
colcon build --symlink-install
source install/setup.bash

3. Run agro_robot_sim (simulation and navigation)

Start full farm simulation/navigation launch:

source /opt/ros/humble/setup.bash
source /home/joaodemoura/agro_robot_ws/install/setup.bash
ros2 launch agro_robot_sim fazenda_completa.launch.py

Optional SLAM mode:

ros2 launch agro_robot_sim fazenda_completa.launch.py slam:=True

4. Run caatinga_vision (inference and analytics)

Start the IA pipeline:

source /opt/ros/humble/setup.bash
source /home/joaodemoura/agro_robot_ws/install/setup.bash
ros2 launch caatinga_vision ia_pipeline.launch.py model_path:=~/models/best.pt

Photo capture pipeline:

ros2 launch caatinga_vision photo_capture.launch.py session_id:=sess_demo

Required External Assets

This repository intentionally excludes large artifacts from version control.

You must provide locally:

• Dataset metadata path: datasets/agro_v1/data.yaml (already included)
• Dataset images/labels in local storage when training
• YOLO model weights, for example: ~/models/best.pt

Demo and Results

Use this section to attach media when showcasing the project:

• Simulation + navigation demo GIF placeholder: docs/media/sim_navigation.gif
• Vision inference overlay screenshot placeholder: docs/media/inference_overlay.png
• Analytics/traceability output screenshot placeholder: docs/media/traceability_status.png

How to generate media:

1. Run simulation and IA pipelines together.
2. Record screen using OBS or GNOME Screen Recorder.
3. Export short GIFs/screenshots and save them under docs/media/.
4. Update this README with embedded markdown images.

Limitations

• Full field validation with physical hardware is not part of this repository.
• Runtime quality depends on camera conditions and YOLO model quality.
• Some launch files are tailored to ROS 2 Humble and current package versions.

Next Steps

• Add repeatable benchmark scenarios for navigation and detection quality.
• Add automated unit/integration tests for critical nodes.
• Publish curated demo media and performance metrics per release.

Contact

• GitHub: joaodemouragy-hash
• Email: joaodemouragy@gmail.com