A feedforward neural network wheelchair driving joystick

People with disabilities such as those affected by IMC disease or Parkinson´s disease have difficulties in operating standard joystick due to their different levels of tremors or difficulties encountered when moving their arms. The objective of this work is to design a new neural joystick suitable for each patient allowing overcoming these difficulties or incomplete or erroneous actions, in order to reach maximum security. The design of the neural network joystick system is based on training a neural network to model the inverse of the standard joystick. The trained resulting neural network is then connected to output of the joystick. The overall system is then used to control all the DC motors and devices of the wheelchair. Simulations and experimental real data recorded on disabled persons are then used to highlight the effectiveness of the designed system.