ROS 2 Camera-Based Autonomous Rover

ROS2 Camera-Based Autonomous Rover is a simulation-first mobile robotics project built using ROS 2 (Jazzy) and Gazebo Sim (gz-sim).
The project focuses on vision-based perception, obstacle reasoning, and clean ROS 2 architecture, with an emphasis on explainable autonomy rather than black-box navigation.

This repository is structured to scale from pure visualization → perception → decision-making → control.

---

🌟 Why This Project?

This project was built to:

• Understand camera-based perception pipelines
• Practice clean ROS 2 package separation
• Develop obstacle avoidance without LiDAR
• Visualize what the robot “sees” and “decides”
• Build autonomy incrementally instead of end-to-end black boxes

The rover operates in a simulated environment and detects obstacles purely from RGB camera input, making it suitable for low-cost real-world robots.

---

🚗 Rover Capabilities (Current)

• 🎥 RGB Camera Integration
• 👁️ Camera-Based Obstacle Detection
• 📐 Region-of-Interest (ROI) Processing
• 🧱 Edge-Based Obstacle Reasoning
• 🧠 Left / Center / Right Obstacle Classification
• 🖼️ Real-Time Perception Visualization
• 🧪 Fully Simulated in Gazebo Sim

🚧 Motion control and FSM-based navigation are intentionally not enabled yet to maintain perception clarity.

---

🧠 Perception Pipeline

The obstacle avoidance logic follows a transparent vision pipeline:

1. RGB Image Acquisition
2. ROI Cropping (Front View)
3. Grayscale Conversion
4. Gaussian Blur
5. Canny Edge Detection
6. Spatial Analysis
   • Left / Center / Right regions
7. Obstacle Decision Output
   • LEFT, CENTER, RIGHT, or NONE

This approach prioritizes interpretability and debugging visibility.

---

🧩 System Architecture

The project uses a modular ROS 2 design:

🧱 Simulation

• Gazebo Sim (gz-sim)
• Custom rover URDF/Xacro
• RGB camera sensor

🧠 ROS 2 Packages

• four_wheel_description
  • Robot model
  • Sensors
  • Gazebo integration
• four_control
  • Camera visualization
  • Obstacle perception logic
  • Decision reasoning (no control yet)

👀 Visualization

• OpenCV windows (camera + ROI edges)
• Gazebo Sim GUI
• RViz2 (optional)

This separation ensures:

• Easy debugging
• Clear responsibility boundaries
• Smooth transition to real hardware later

---

🧪 Obstacle Avoidance Logic (Camera-Based)

Obstacle detection is performed using image structure, not distance sensors.

Decision logic:

• CENTER obstacle → highest priority
• LEFT obstacle → free space likely on right
• RIGHT obstacle → free space likely on left
• NONE → clear path

The result is visualized directly on the processed image to ensure decision correctness before control is enabled.

---

🛠️ Tech Stack

• ROS 2 Jazzy
• Gazebo Sim (gz-sim)
• Python (rclpy)
• OpenCV
• cv_bridge
• URDF / Xacro
• RViz2

---

▶️ How to Run

Build the workspace

colcon build
source install/setup.bash

Launch the rover in Gazebo

ros2 launch four_wheel_description rover.launch.py

Run camera-based obstacle visualization

ros2 launch four_control visualization.launch.py