Self-assisted rehabilitation of neurological injury through patient-operated telerobots offers potential benefits over traditional therapy including increased patient involvement, improved timing coordination, and better sensory information. Bimanual self-assist has been clinically shown to demonstrate greater improvements in range of motion and functional recovery than traditional therapy alone. This dissertation generalizes self-assist to lower limb rehabilitation with upper limb assistance, providing a scientific foundation for the mechanisms of self-assist. Through a series of experiments using electromechanical devices under real-time control, we have demonstrated that upper limb involvement enhances lower limb performance, improves anticipation of assistance, and maintains muscle activation.

A critical concern to assistive rehabilitation is the degree to which subjects are capable of coordinating lower limb effort with the assisting effort. In a study involving lower limb disturbance rejection with neurologically intact subjects (n = 12), we found that subjects exhibit improved anticipation and compensation when externally applied loads are self-generated as opposed to generated by an outside agent. The centralized control inherent in self-assist ensures that assisting forces are expected, and even anticipated, compared to assisting forces from an external agent, such as a therapist or computer.

Multi-limb control can represent a significantly increased challenge in motor planning and execution over single limb control. We evaluated the tradeoffs associated with multi-limb control in motor tasks with neurologically intact subjects (n = 7 separated into two experiments). The benefits of adding an upper limb, including increased strength and dexterity, seem to offset any detriments associated with increased control requirements, as evidenced by performance improvements relative to single limb control. In a final motor coordination study, hemiparetic subjects (n = 15) practiced dorsiflexing the impaired ankle with no assistance, upper limb self-assist, computer-assist, and experimenter-assist. We found that any assistance can be used to improve task performance, but self-assist, through improved information and awareness, maintains subject muscle recruitment and effort.