|dc.description.abstract||The purpose of this thesis was to design, test, and evaluate a three degree of freedom, wheelchair-mounted, robotic upper limb exoskeleton that detects human intent for actuation. The exoskeleton was designed in SolidWorks, followed by a finite element analysis to determine stress and displacement under average human arm loading. A workspace analysis was performed using the Denavit-Hartenberg method in MATLAB to determine the effective reach of the exoskeleton end-effector. The exoskeleton was printed with out of PLA and NylonX plastics with a RAISE3D Pro2 Plus printer. This exoskeleton was actuated with three NEMA 17 stepper motors which were coupled with encoders at each joint to provide safe rotation within biomechanical limitations. Force myography and electromyography were utilized in the controller to detect human intent. For evaluation, the exoskeleton was tested on 10 able-bodied human subjects from 18-60 years old with IRB approval.
FEA maximum stress expected is 429.1 MPA. For FMG individual joint tests, shoulder flexion/extension had the highest average desired movement at 67%, then elbow flexion/extension at 61%, and shoulder external/internal rotation at 36%. EMG neural network accuracies were very high, with the minimum overall neural network accuracy being 87.7% and the maximum reaching 97.4%. For EMG controller individual joint testing, shoulder external/internal rotation was highest at 72%, shoulder flexion/extension at 68%, and elbow flexion/extension at 50%. For the waterbottle pick and place, the FMG control times ranged from 14.9 to 82.7 seconds, while EMG control times ranged from 15.3 to 62.6 seconds. Both control systems worked fairly well, though shoulder external/internal rotations show a need for improvement for both FMG and EMG systems. Future work includes PID control, metal links, and testing on subjects with cerebral palsy and muscular dystrophy.||