Autonomous Wheelchair Navigation

Student thesis: Master thesis (including HD thesis)

  • Søren Lynge Pedersen
  • Jeppe Møller Holm
This Project describes how to make an autonomous wheelchair for people with Amyotrophic Lateral Sclerosis (ALS). Because people with ALS cannot use their muscles to propel or even control a nor- mal wheelchair, it is necessary that the wheelchair can be controlled by the only part of the body that they still have full control over; their eyes. It is investigated how to sense the house, how to localise the wheelchair in the house, how to plan a path through the house and how to drive a wheelchair. Mounted on the wheelchair to sense the house is a laser range scanner. The output from this along with the odometry is used in a particle filter for localisation, incorporat- ing an Augmented Adaptive Monte Carlo Localisation. To plan the path through the house, the Configuration Space of the house is com- puted and a Generalised Voronoi Diagram calculates waypoints with the maximum distance to the walls. A Dijkstras Algorithm finds the shortest path from start to goal. To avoid unmapped obstacles, repulsive velocity vector fields are implemented. The controller to control the wheelchair is an optimal model based LQR. This project has been a proof of concept, where it has been shown that it is pos- sible to make an autonomous wheelchair, capable of transporting a user with Amyotrophic Lateral Sclerosis, or any other disabled per- son, with no user input except where to go. The wheelchair does not need to be told where it is, or how to get to the goal. It does not even need any artificial landmarks for navigation. Using only on- board sensors and a map of the house, which was available anyway, it performs the required tasks.
Publication date2008
Number of pages120
Publishing institutionSection for Automation and Control
ID: 14412617