A neuro-fuzzy approach to mitigate tremors due to multiple-sclerosis effects in a wheelchair joystick controller

Ongoing research efforts in the field of Artificial Intelligence aim at providing electromechanical systems with more intelligence so that they can operate autonomously. Advancing in the same direction, this thesis proposes a Neuro-Fuzzy based joystick controller design for developing easy navigation techniques for a navigating wheelchair. Unfortunately, many wheelchair users suffer from Multiple Sclerosis causing tremors, which leads to incorrect control signals to the joystick. The main aim of this research is to report the design and simulation of a controller for an electric wheelchair to minimize tremors as well as adapt to a patient's peculiar controlling commands using Neuro-Fuzzy techniques. The controls of an electronic wheelchair are designed to execute the directional commands given by the user. The purpose of the fuzzy controller is to modify the wheelchair user's erratic joystick movements to control the wheelchair smoothly and accurately. The learning and adaptation system is designed using Pseudo-Outer Product neural learning to modify the membership functions of the fuzzy logic control system for optimizing the controller to the tremor characteristics of a specific user.