The design and evaluation of an interface and control system for a scariculated rehabilitation robot arm

This thesis is concerned with the design and development of a prototype implementation of a Rehabilitation Robotic manipulator based on a novel kinematic configuration. The initial aim of the research was to identify appropriate design criteria for the design of a user interface and control system, and for the subsequent evaluation of the manipulator prototype. This led to a review of the field of rehabilitation robotics, focusing on user evaluations of existing systems.
The review showed that the design objectives of individual projects were often contradictory, and that a requirement existed for a more general and complete set of design criteria. These were identified through an analysis of the strengths and weaknesses of existing systems, including an
assessment of manipulator performances, commercial success and user feedback.
The resulting criteria were used for the design and development of a novel interface and control system for the Middlesex Manipulator - the novel scariculated robotic system. A highly modular architecture was adopted, allowing the manipulator to provide a level of adaptability not approached by existing rehabilitation robotic systems. This allowed the interface to be configured to match the controlling ability and input device selections of individual users.
A range of input devices was employed, offering variation in communication mode and bandwidth. These included a commercial voice recognition system, and a novel gesture
recognition device. The later was designed using electrolytic tilt sensors, the outputs of which were encoded by artificial neural networks. These allowed for control of the manipulator through head or hand gestures.
An individual with spinal-cord injury undertook a single-subject user evaluation of the Middlesex Manipulator over a period of four months. The evaluation provided evidence for the
value of adaptability presented by the user interface. It was also shown that the prototype did not currently confonn to all the design criteria, but allowed for the identification of areas for design improvements.
This work led to a second research objective, concerned with the problem of configuring an adaptable user interface for a specific individual. A novel form of task analysis is presented
within the thesis, that allows the relative usability of interface configurations to be predicted based upon individual user and input device characteristics. An experiment was undertaken with 6 subjects performing 72 tasks runs with 2 interface configurations controlled by user gestures. Task completion times fell within the range predicted, where the range was generated using confidence intervals (α = 0.05) on point estimates of user and device characteristics. This allowed successful prediction over all task runs of the relative task completion times of interface configurations for a given user.