Toward low-cost only vision autonomous wheelchair for disabled people: Corridor trial. (English)
We aim to improve the lives of disabled people. This project is motivated by the needs of Syrians whose mobility disabilities, many resulting from the war, will require assistive wheelchairs and the support of others for the rest of their lives. We believe that the development and use of new technologies to assist them will have a major positive impact on their lives. Our study sought to develop a cheap semi-autonomous visual navigation-powered-wheelchair system, using off-the-shelf computing devices and imaging technology. This will help to make electric wheelchairs smarter and cheaper, as well as more accessible to those with mobility disabilities, not just in Syria, but across developing countries. Encouraged by the growing number of successful machine-learning applications, we developed and tested a novel machine-learning wheelchair-navigation system to enable electric wheelchair users with minimal mobility to avoid obstacles and risk, increasing their safety and mobility. It is the first to use Gaussian Processes (GPs) as a machine learning approach for a wheelchair task and experiment results show superior capabilities and performance, as compared to existing variations.