As modern medicine has improved, the average age of patients has increased. This has cause a growing number of patients to develop disabilities over time due to spinal cord injuries and stroke among other neurological ailments. This has led to an increased interest in developing robotic exoskeletons to help patients with neuromuscular rehabilitation. However, most exoskeletons do not accurately replicate the natural human gait kinematics due to a lack of degrees of freedom at the designed knee joint. In this thesis, the leg assembly for a robotic rehabilitation (RoboREHAB) device is redesigned to improve the gait kinematics and a reinforcement learning (RL) based controller is designed to control the new leg assembly using motion capture data. The new leg assembly and RL controller performed with a 5% margin of error from the motion capture data. Further improvements will be made to construct a full-scale prototype and establish real-time data acquisition.

Date of publication

Spring 5-15-2024

Document Type




Persistent identifier


Committee members

Chung-Hyun Goh, Alwathiqbellah Ibrahim, Mohammad Biswas, Mukul Shirvaikar


Master of Science in Mechanical Engineering