Drug Delivery Robot

Introduction

This is a robot that can be used to deliver drugs to patients in a hospital. The robot will drive through the sparse cones and follow the nurse in the end of the road.

The function implemented is not complicated. However, it is a good pratice of how to integrate YOLOv5 (or other deep learning models) into ROS and how to use Python to write ROS packages.

Features

• Crafted entirely in Python.

• Integrate YOLOv5 into ROS and used it to detect the cones in the image published by the camera.

• Color mask implemented in OpenCV is used to detect the nurse.

Installation

Prerequisites

• Ubuntu 20.04

• ROS Noetic

• Python 3.8 (Comes with ROS Noetic installation)

• zsh (It is recommended to use zsh as it has better auto-completion than bash)

It is recommended to have a basic understanding of ROS Topics. The following tutorial will walk you through the process of writing a simple publisher and subscriber.

• Writing a Simple Publisher and Subscriber (Python)

Dependencies

Note

A virtual environment is not recommended for ROS as it does not officially support this feature. All python packages should be installed globally using pip install. If you cannot find pip, you can install it by running:

sudo apt install python3-pip

YOLOv5

You can install YOLOv5 by running the following the instructions in the YOLOv5 repository

If you are using a custom model, you need to modify the 'listener.py' file to load your model.

model = torch.hub.load('ultralytics/yolov5', 'custom', path='path/to/best.pt') # local model
# or
model = torch.hub.load('path/to/yolov5', 'custom', path='path/to/best.pt', source='local')  # local repo

OpenCV

To use OpenCV in ROS, you need to refer to this tutorial to install OpenCV.

Note that 'opencv2' is not required to be included in the find_package command in CMakeLists.txt as it is already included in cv_bridge.

PyTorch and other python packages

They should be installed with ease using pip install.

Usage

This is a ROS package. To use it, you need to clone this repository to your catkin workspace and run catkin_make.

For example, if your catkin workspace is ~/catkin_ws, you can run the following commands to clone this repository and build it:

cd ~/catkin_ws/src # Go to your catkin workspace
catkin_create_pkg your_package_name # Create a new package
cd your_package_name # Go to your new package
rm rf * # Remove all files in the package
git clone https://github.com/kowyo/drug-delivery-robot
cd ~/catkin_ws # Go back to your catkin workspace
catkin_make # Build the package

After building the package, you can run the following command to launch the robot:

To launch the camera and publish the image

cd ~/catkin_ws
source devel/setup.zsh
rosrun your_package_name talker.py

To subscribe to the image and detect the cones

cd ~/catkin_ws
source devel/setup.zsh
rosrun your_package_name listener.py

Demo

292152300-c0b99b8b-1e8a-487d-aca7-a1e24230e809.MOV

Citation

If you use this repository in your research, please cite our work as follows:

@software{Drug Delivery Robot,
  title = {Drug Delivery Root},
  author = {Zifeng Huang, Shunyu Zhou},
  year = {2023},
  url = {https://github.com/kowyo/drug-delivery-robot}
}