Individuals with severe motor impairments often have a difficult time performing daily activities, and thus rely heavily on assistive devices to restore some functional independence. The two main limitations of the current controllers for assistive devices are: 1) controllers that do not require coordinated hand movement provide a limited vocabulary of commands, and 2) controllers do not promote engagement of remaining mobility. Devices like the sip-n-puff or head array do not provide full continuous control, but instead allow the user to deliver a series of one-dimensional commands at one of only a few predetermined speeds. Further, these devices do not encourage users to practice and reinforce what residual movements remain post injury. Together, these issues represent a gap in the development of assistive device controllers. The body-machine interface (BMI) was designed to address these issues. Motor learning studies suggest that redundancy is a key aspect of how we move, providing flexibility and performance benefits for everyday movements. The BMI leverages this to allow individuals with severe motor impairments to control assistive devices while also encouraging them to utilize their remaining mobility. Previous results have demonstrated that people, including individuals with high level spinal cord injuries, can learn to reorganize high dimensional movements to control a computer cursor using a BMI. I first developed and validated a system that uses the BMI to control a power wheelchair. Three individuals with cervical spinal cord injuries learned how to use small shoulder movements, measured with inertial measurement unites (IMUs), to control the speed and direction of a power wheelchair. After some training, all subjects were able to accurately maneuver the power wheelchair through an obstacle course without any collisions. Additionally, performance using the BMI was comparable to performance using a joystick, the preferred method of control for all participants. This validated that subjects could learn to reorganize shoulder movements to skillfully control a power wheelchair. Next, I attempted to reverse engineer principles of optimal control to encourage subjects to learn specific movements when using a BMI. Many studies have suggested that people learn the specific movements that minimize the direct effect of signal dependent noise on task performance. To exploit this idea, I added a specific pattern of signal dependent noise between the execution of a movement and its task consequences. Through training, subjects began to make movements that were biased towards the specific ones that minimized the effects of noise. This not only provides particularly direct evidence that people actively seek to minimize noise in their movements, but this paradigm also provides a method to design noise to teach specific control strategies that can provide functional benefits for users. Finally, I sought to strengthen the noise paradigm by classifying how subjects react to changing patterns of noise. In the second study, I showed that noise could bias the initial movements people learn, however it was unclear how robust these biases were to changes in the pattern of noise. I found that not only do people learn movements that minimize noise during initial exposure to a task, they continuously update the movements they make in response to changes in noise parameters. This provides some evidence to further clarify how people perceive and process noise in redundant tasks, while also strengthening the possibility to use this approach to provide rehabilitative benefits, as the noise could be changed as rehabilitation progresses to continually address changing movement issues

Last modified

• 04/27/2018

Creator

• Elias Thorp

DOI

• https://doi.org/10.21985/N2FX2M

Subject

• Biomedical Engineering

Keyword

• Human machine interface
• Redundancy
• Motor learning
• Power wheelchair
• Rehabilitation
• Motor Noise

Date created

• 2017-01-01

Resource type

• Dissertation

Rights statement

• In Copyright