Advanced semi-autonomous wheelchair system for people with disabilities
- Publication Type:
- Issue Date:
Powered wheelchairs have been dedicated to providing independent mobility to individuals who have physical or cognitive impairments and disabilities. Unfortunately, there are still a significant number of impaired people who find it difficult or impossible to use the standard powered wheelchairs due to their high level of injuries. Intelligent wheelchairs, therefore, have been developed to cater for these users, utilizing control methods and assistive navigation algorithms to help the patients safely navigate an environment. This thesis aims at developing an intelligent wheelchair that provides functional mobility for the disabled users. Unlike a number of intelligent wheelchairs based on algorithms from the robotic field, this dissertation investigates two advanced techniques, which are Bayesian theory and artificial neural network, in building the semi-autonomous navigation system for the intelligent wheelchair. In particular, Bayesian theory is employed to determine free spaces for navigation and make adaptive shared control decisions between the users and the machine. In the meantime, the artificial neural network is recruited to learn the navigation skills through a number of patterns. During the training, the Bayesian framework is also applied to optimize the network structure as well as the weight values. After training, the network is able to navigate the wheelchair into the location of interest without colliding with surrounding obstacles. The developed semi-autonomous navigation system was experimentally verified on the real wheelchair with a number of participants. The experimental results convincingly demonstrate that our system is able to real - time navigate in the unknown environment with safe, smooth and optimal trajectories, and that it can be a valid alternative for the people with disabilities.
Please use this identifier to cite or link to this item: