Ackermann Mobile Robot: Perception-to-Action Pipeline

This repository contains a complete ROS 2 perception-to-action pipeline for an Ackermann-steered mobile robot, developed for the Greenswip Robotics Software Assignment. The system enables a robot to autonomously identify, track, and navigate toward a specific red box target while ignoring various decoy objects in a Gazebo Harmonic simulation.

🚀 Project Overview

The project integrates computer vision and advanced control theory to solve a "Target Retrieval" task under strict kinematic constraints.

• Robot Kinematics: Ackermann Steering (No in-place rotation).
• Simulation Environment: Gazebo Harmonic (ROS 2 Jazzy).
• Perception: OpenCV-based HSV color segmentation and contour analysis.
• Control: Proportional control mapped to Ackermann steering geometry.

🛠️ Technical Implementation

Architecture & Setup

The robot is described via a URDF (ack.urdf.xacro) enhanced with Gazebo Sim plugins. The communication between ROS 2 and Gazebo is handled by the ros_gz_bridge.

Perception (Computer Vision)

The perception_node processes a live camera feed. It utilizes:

• HSV Masking: Specifically tuned to isolate the Red Box (ignoring blue, orange, and green decoys).
• Centroid Tracking: Calculates the horizontal error from the image center to provide feedback for the steering controller.

Navigation & Control

The control_node implements an Ackermann-specific navigation algorithm:

• Kinematic Constraints: The robot maintains a constant forward velocity while turning, ensuring it follows a curved path rather than pivoting, satisfying strict Ackermann constraints.
• Closing Logic: The robot automatically decelerates and halts once it reaches a proximity threshold determined by the target's visual area.

📂 Repository Structure

ros2_ws/
└── src/
    └── ackermann_nav/
        ├── ackermann_nav/      # Perception and Control Nodes
        ├── launch/             # ROS 2 Launch files
        ├── urdf/               # Robot URDF with GZ plugins
        └── worlds/             # Gazebo World with shapes

📺 Video Demonstrations

Below are the demonstrations of the robot successfully reaching the target in three different shuffled object arrangements.

Arrangement 1 (Center)

center_box.mp4

Arrangement 2 (Right Side)

right_side_box.mp4

Arrangement 3(Left Side)

left_side_box.mp4

🏁 How to Run

1. Build the Workspace

In your WSL2 terminal:

cd ros2_ws
source /opt/ros/jazzy/setup.bash
colcon build --symlink-install

2. Launch the Simulation

source install/setup.bash
ros2 launch ackermann_nav main.launch.py