Date of Award

Winter 1-1-2019

Degree Type

Dissertation

Degree Name

Doctor of Philosophy In Engineering

Department

Mechanical Engineering

First Advisor

Richter, Hanz

Second Advisor

Antonie J. van den Bogert

Third Advisor

Eric Schearer

Abstract

Human-machine interaction has become an important area of research as progress is made in the fields of rehabilitation robotics, powered prostheses, and advanced exercise machines. Adding to the advances in this area, a novel controller for a powered transfemoral prosthesis is introduced that requires limited tuning and explicitly considers energy regeneration. Results from a trial conducted with an individual with an amputation show self-powering operation for the prosthesis while concurrently attaining basic gait fidelity across varied walking speeds. Experience in prosthesis development revealed that, though every effort is made to ensure the safety of the human subject, limited testing of such devices prior to human trials can be completed in the current research environment. Two complementary alternatives are developed to fill that gap. First, the feasibility of implementing impulse-momentum sliding mode control on a robot that can physically replace a human with a transfemoral amputation to emulate weight-bearing for initial prototype walking tests is established. Second, a more general human simulation approach is proposed that can be used in any of the aforementioned human-machine interaction fields. Seeking this general human simulation method, a unique pair of solutions for simulating a Hill muscle-actuated linkage system is formulated. These include using the Lyapunov-based backstepping control method to generate a closed-loop tracking simulation and, motivated by limitations observed in backstepping, an optimal control solver based on differential flatness and sum of squares polynomials in support of receding horizon controlled (e.g. model predictive control) or open-loop simulations. v The backstepping framework provides insight into muscle redundancy resolution. The optimal control framework uses this insight to produce a computationally efficient approach to musculoskeletal system modeling. A simulation of a human arm is evaluated in both structures. Strong tracking performance is achieved in the backstepping case. An exercise optimization application using the optimal control solver showcases the computational benefits of the solver and reveals the feasibility of finding trajectories for human-exercise machine interaction that can isolate a muscle of interest for strengthening.

Share

COinS