ROS Line Following Robot

This project is a small line-following robot. It can detect and follow tape lines on my floor! The robot runs ROS on a Raspberry Pi, uses OpenCV to detect lines, and an Arduino Pro Micro to control differential steering.

I built this small project to sharpen my ROS and OpenCV skills. Your mileage may vary; the OpenCV code likely will not work in a different lighting environment. The perception system won't generalize since it relies on the line being significantly lighter than the floor, which works in my workspace.

Check out the robot in action at https://www.youtube.com/watch?v=0DcV46YkHR8.

Architecture

┌────────────────────────────────┐ ┌─────────────────────────────────────────────────────────┐
│ ROBOT HARDWARE                 │ │ ROS NODES                                               │
│ ┌──────────────┐ Raw           │ │ ┌────────────────────────┐ Processed ┌────────────────┐ │
│ │ Raspberry Pi │ Image         │ │ │ Image Preparation      │ Image     │ Line Detection │ │
│ │    Camera    ├───────────────┼─┼─► Threshold,             ├───────────► Path Planning  │ │
│ └──────────────┘               │ │ │ Perspective-Correction │           └───────┬────────┘ │
│                                │ │ └────────────────────────┘                   │          │
│ ┌──────────────┐ Speed/Angle   │ │                                              │          │
│ │ Arduino Pro  │ Command       │ │ ┌────────────────────────┐  Speed/Angle      │          │
│ │ Micro        ◄───────────────┼─┼─┤ Serial Communication   │  Command          │          │
│ └──────┬───────┘               │ │ │ Node                   ◄───────────────────┘          │
│        │ Motor                 │ │ └────────────────────────┘                              │
│        │ Command               │ │                                                         │
│ ┌──────▼─────────────────┐     │ └─────────────────────────────────────────────────────────┘
│ │       Dual Motor       │     │
│ │       Controller       │     │
│ └─────┬────────────┬─────┘     │
│       │            │           │
│ ┌─────▼────┐ ┌─────▼─────┐     │
│ │Left Motor│ │Right Motor│     │
│ └──────────┘ └───────────┘     │
│                                │
└────────────────────────────────┘

The robot architecture is a unidirectional pipeline. First, sensor data is fed from the Raspberry Pi camera, and through image processing which performs a perspective correction and contrast/threshold image adjustment to facilitate line detection.

Once the image is processed, a line detection algorithm is run to find the approximate direction of the line to follow. The code uses the relative position and angle of the line to compute a target speed/angle that will track the line.

The control data is sent over a serial connection to an Arduino Pro Micro, which computes individual motor commands and commands a dual motor controller.

Development Notes

Prerequisites

To run this project, you must have a ROS Noetic installation setup on a Raspberry Pi. This repository should be checked out into a Catkin workspace. For more information, see the ROS Tutorials.

You must also have the following dependencies installed:

• Python 3 (sudo apt install python3.8 && sudo apt install python3-pip)
• OpenCV (pip install opencv-python)
• CvBridge (sudo apt install ros-noetic-cv-bridge)
• PySerial (pip install pyserial)
• numpy (pip install numpy)
• rosserial_python (sudo apt install ros-noetic-rosserial-python)
• robot_upstart (sudo apt install ros-noetic-robot-upstart)
• video_stream_opencv (sudo apt install ros-noetic-video-stream-opencv)
• raspicam_node (See https://github.com/UbiquityRobotics/raspicam_node for build/install instructions)

Starting the Project

To run the entire project, use the roslaunch file in this repository: roslaunch line_follower line_follower.launch

Remote ROS

To connect to the remote ROS instance, set the following environment variables:

export ROS_MASTER_URI=http://linefollower:11311
export ROS_IP=linefollower

Start the visualization:

rosrun rviz rviz -d ~/linefollower/catkin_ws/src/ros-line-follower-robot/rviz/line_follower.rviz

Starting/Stopping the Background Service

The entire stack can be launched in the background using robot_upstart.

To install the background service:

rosrun 
    robot_upstart install 
    --job linefollower
    --setup /home/ubuntu/catkin_ws/devel/setup.sh
    --logdir /var/log/upstart
    --symlink line_follower/launch/line_follower.launch
sudo systemctl daemon-reload

To start the background service:

sudo systemctl enable linefollower
sudo systemctl start linefollower

To stop the background service:

sudo systemctl stop linefollower
sudo systemctl disable linefollower

Topics

• /camera/raspicam_node/image: An image from the Raspberry Pi camera.
• /camera/raspicam_node/image: A mockup video published on a loop, used to test the CV pipeline.
• /line_follower/processed_image: The processed image: threshold, reverse perspective transform.
• /motor_driver/speed: A commanded speed for the Arduino to achieve.
• /motor_driver/direction: A commanded direction for the Arduino to achieve.