This Is AuburnElectronic Theses and Dissertations

Sensing, Classifying Movements, and Actuation of Biomimetic Robots




Kennedy, Scott

Type of Degree

Master's Thesis


Mechanical Engineering


Human skeletal muscle motion is a complex electrical, chemical, and mechanical process. The electrical, or electromyography, signal originates from the brain and travels to the muscle initiating the contraction of the muscle. This signal can be detected in the muscle roughly 500 ms before movement occurs and can be used to classify the motion of the subject to inform the motion of an assistive robot. The electromyography signal can be monitored by surface electromyography sensors adhered to the subject's skin. These sensors are highly dependent on position and orientation of the sensor with respect to the belly of the muscle being monitored and requires both knowledge of musculature and extensive setup time to use. In order to implement a better sensor solution capable of being applied without any knowledge of anatomy, a textile suit was outfitted with electromyography sensors. The suit was shown to accurately classify the motion of the user using a K-nearest neighbor algorithm compared to the traditional method of sensor adhesion. This suit works to inform the movement of an exoskeleton. Along with detecting intent of motion from the user, exoskeletons and assistive robots need to move like their users. Shape memory alloy actuators have previously been used as micro-actuators. With a high strength to weight ratio, small form factor, and low cost, shape memory alloys are an attractive actuator option that contracts and lengthens similarly to human skeletal muscles. However, shape memory alloy actuators also have high power consumption, low strain (4-8\%), and low operating frequency ($<$ 3 Hz). In order to overcome these drawbacks, a shape memory alloy actuator was designed and tested in a bimorph configuration. This shape memory alloy actuator design addressed the drawbacks of traditional shape memory alloy actuators and could potentially expand the usage of this technology in biomimetic robots.