Development of a Lower Limb Exoskeleton and Application of Learning from Demonstrations for Controllers

Goldfarb, Nathaniel

Etd

Development of a Lower Limb Exoskeleton and Application of Learning from Demonstrations for Controllers

Public

Robotic lower limb exoskeletons play a crucial role in the future of rehabilitation for people with spinal cord injuries and other physical impairments. The research field of lower limb exoskeletons is vast and made up of several distinct and connected fields. This dissertation examines methods to test and develop new robotic lower limb exoskeletons and controllers. Several new exoskeleton knee joint designs are presented and tested with customizable and reconfigurable exoskeletons. A simulation platform is presented to test both controllers using an open-source simulation package. Human motion data is collected, parsed, and encoded to build models for generalized trajectories; this is shown for both stair climbing and gait cycles reproduction. Two different controller architectures are also presented. The first is a new approach to using an iterative linear quadratic regulator with learning from demonstration. The second controller is an assistive controller for shared control between a person and exoskeleton. This assistive controller expands on prior literature to incorporate two closed-loop controllers. The analysis presented shows the response of the assistive controller at different engagements and interaction levels.

Creator