Autonomous Wheelchair

The purpose of this project was to develop a wheelchair for children affected by cerebral palsy. The wheelchair is capable of navigating across a structured environment like a house, hospital etc. A LiDAR sensor is used for navigation, localisation and mapping. Sampling based algorithm like RRT and RRT* have been implemented to provide robust path planning.

The project was executed with AGILE sprint planning methodology. The tasks were recorded for the time taken, and the commits to the GitHub repository were timely and constructive. The GitHub repository of this project is hosted here

The project was split up into 5 main sections:

Control - Controlling the system to follow the differential drive mechanism
Mapping - Mapping a new environment
Odometry - Laser based odometry for accurate localization of the robot
AMCL - Localisation using Adaptive Monte Carlo Localization
Planning - Using the map developed in the previous stages to plan a path using sampling based algorithm

The wheelchair is powered by 2 DC motors, and the front 2 wheels are free rotating caster wheels. A differential drive is thereby developed.

The following videos show the hardware implementation and simulation of the wheelchair.

Mapping operation:

The mapping was done using YDLIDAR X4. The Gmapping package was used in ROS. The wheelchair was teleoperated using ROS Multi-Machine concept. The slave computer was placed on the wheelchair. It was connected to an Arduino Mega, which was used to drive the motors. The master computer would send the velocity information to the slave computer, and it was then mapped accordingly by the arduino onto the motor driver.

The map being developed was visible on the master computer.

Navigation operation:

The wheelchair was simulated in Gazebo. 2 cases were simulated; straight line motion and cross room motion. The simulation videos are as follows:

Detailed explanation of the development of the wheelchair is available in the paper below.