Shared user-computer control of a robotic wheelchair system

An assistive robotic wheelchair system should allow its user to travel more efficiently and with greater ease. The robotic wheelchair system described in this thesis, Wheelesley, performs semi-autonomous navigation for its user, taking high-level directional commands and performing the low-level navigation required to avoid obstacles and stay on a safe path. The system consists of a standard electric wheelchair with an on-board computer for motor control, a vision system running on an off-the-shelf notebook computer, sensors, and a graphical user interface running on a tray mounted notebook computer. Most other research on robotic wheelchairs only addresses indoor navigation. The Wheelesley system can travel both indoors and outdoors using specialized navigation modes; there is a control algorithm for indoor navigation and a control algorithm for outdoor navigation. User tests have been conducted for both control modes to compare robotic assisted control against manual control. Robotic assisted control requires 71% less effort for indoor navigation and 88% less effort for outdoor navigation. In addition, the total time needed to travel between two points is reduced since less time is spent waiting to scan to the desired commands. The system switches automatically between navigation modes through the use of a novel indoor/outdoor detector. The detector is comprised of an ultrasonic transducer, three light-to-voltage optical sensors and a thermistor. A decision tree learned by C4.5 using data collected in a variety of indoor and outdoor conditions classified a test set correctly 98.3% of the time. The system can be easily customized for the access method(s) required by each user. This thesis describes customization of the user interface for two different access methods: eye tracking, an uncommon access method for a wheelchair, and single switch scanning, which is considered the driving method of last resort on standard powered wheelchairs. The wheelchair system and its interface was evaluated by a group of physical therapists. Thesis Supervisor: Rodney A. Brooks Title: Professor Shared User-Computer Control of a Robotic Wheelchair