A self-contained, intelligent, micro-controller expert system to augment powered wheelchair users

• Main Author: Stott, Ian James
• Other Authors: Sanders, David Adrian ; Tewkesbury, Giles Eric
• Published: University of Portsmouth 2002
• Subjects: 629.892; Autonomous
• Online Access: https://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.247093

An innovative prototype intelligent wheelchair was created to test novel algorithms to help disabled wheelchair users navigate through an unstructured environment. Local planning methods were selected to assist in steering. a simple, on-board expert system attempted to find a suitable trajectory that was close to the route requested but which moved away from close objects. The algorithms were weighted by the wishes of the user. If the expert system detected that the user was unsure or inconsistent, the on-board expert was given more importance in the selection of the wheelchair route. The expert system could be over-ridden by the user if the user was consistent in the use of the joystick. The prototype intelligent wheelchair could detect the environment, modify wheelchair control data and detect the wishes of the user. Decisions were made by the expert system and pre-planned responses could be activated. Autonomous operation of the prototype intelligent wheelchair was demonstrated during tests. A micro-controller was embedded into the wheelchair control path. Information was read from the joystick and new sensor system and signals were sent to the wheelchair controller. The raw sensor data were processed to improve the reliability of the range data by mapping the sensor data onto a histogram certainty grid. The intelligent wheelchair created during this work augmented the control that the disabled user of the wheelchair could provide. It was important that the expert system operated in real time in order to assist the user. There were two real time inputs; the user input and the sensors. The user indicated a speed and direction for the wheelchair. The sensors gathered information about the environment. A <i>sensor expert </i>system then analysed the sensor information and made a recommendation for a path that would prevent collisions. The data inputs sometimes conflicted. Another expert, called <i>Fuzzy Mixer</i> considered both inputs and was responsible for the final outputs to the motor controller. the <i>joystick monitor </i>expert was responsible for interpreting the wishes of the user. Variables such as the joystick position and consistency were examined by the <i>joystick monitor  </i>to assess the desired wheelchair trajectory. The prototype intelligent wheelchair was tested in the laboratory. It was self-contained which allowed realistic testing without the burden or restriction of trailing umbilical cables. Obstacles were placed in the path of the wheelchair and the response was encouraging.