Indoor location and collision feedback for a powered wheelchair system using machine learning

In 2015 a powered wheelchair system to detect and avoid objects was enhanced with a Raspberry Pi to extend the sensory input sources to the system in order to record information about its environment. Wheelchair users are not always able to use simple controls such as joysticks to drive and they may have to control the wheelchair using their tongue, head or feet. This can increase the effort it takes to learn how to drive and therefore it becomes important to track and observe how a wheelchair user is progressing. The research described in this paper employs machine learning to use wireless access points and predict its location, and with prolonged use will learn routes between rooms and buildings. The system uses location and accelerometer data to present information about driving patterns and collisions behaviour. The primary user interface is for the wheelchair user to orientate themselves indoors whilst driving and a secondary user interface is displaying past information to a carer to inform about incidents and general tracking of the wheelchair user.