Autonomous Electric Wheelchair

Computer vision has many applications in robotics. This project aims to use computer vision as a feedback loop for an electric wheelchair . Allowing it to navigate the indoor enviroment. At this point the project has developed a mechanism for Remote Control of the wheelchair over a wireless network with control achievable through a web browser.

Contorl and Response signals can be sent and received from the wheelchair with a GET function written in PHP. This has demonstrated a high-speed respnse to commands. Kinect Camera Streams for both the RGB and Depth Image are also avalible for processing through an MJPEG stream. This allows computationally intesive image processing and navigation decisions to be made remotely reducing power consumption at the wheelchair itself.

System

Dependencies

This project is based largely on interfacing with a physcial hardware, such as motor, relays, etc. As a result a list of hardware dependencies has also been included.

Software Dependencies

For manual control via web browser the following dependencies are required;

OpenCV-4.0.0-alpha (https://github.com/opencv)
OpenKinect Library (https://github.com/OpenKinect/libfreenect)
MJPG Streamer (https://github.com/jacksonliam/mjpg-streamer)
Apache
PHP 7.0 or Higher

For autonomous navigation the above dependencies and the follow are requird;

Anaconda Python Platform (https://anaconda.com) with the following packages installed
Numpy conda install -c anaconda numpy
Opencv conda install -c conda-forge opencv
Matlibplot conda install -c conda-forge matplotlib
pip conda install pip
imutils pip install imutils

Hardware Dependencies

Arduino MEGA
RaspberryPi 3
Microsoft Kinect V1 (Note that the V2 Kinect Sensor requires USB 3.0 Support)
Pololu G2 H-Bridge Motor Driver (24v21)

Installation

Installation of Apache (Any web-server will do, but I prefer Apache) iincluding PHP. This guide from howtoraspberrypi is helpful for beginners (https://howtoraspberrypi.com/how-to-install-web-server-raspberry-pi-lamp/).
Build and Install OpenCV. I used 4.0.0-alpha. However, you can use the most up to date version. This tutorial will get you started at pyimagesearch
Clone, Build and Install the OpenKinect library. Here is a great tutorial for doing just that.
Copy html folder to webserver folder. Appropriate permissions are required to run PHP and Python Scripts for the interface

Acknowledgements

Queen’s University Belfast
PyImageSearch