| dc.description.abstract | A powered transfemoral prosthesis is an assistive device for patients with above knee
amputation. For these patients, walking on various sloped surfaces is one of the most challenging
tasks in their daily lives. Designing prostheses that can effectively adapt to varying
terrain remains an ordeal to this date. In this thesis, we focused on generating the desired
trajectories for various inclined surfaces without prior knowledge of slopes using human
impedance and cubic-Bezier-polynomials-based optimization. Trajectory generation for
the powered prosthesis is an important procedure to design an appropriate controller that
mimics human locomotion; the trajectory has to be generated for each gait cycle in realtime
to produce a stable, robust, and human-like walking. The proposed method is rooted
in analyzing the human data from the motion capture system, to gain an understanding
of how human walks differently according to the slopes. Impedance control using human
parameters allows the prosthesis adapt to these different slopes during the stance phase.
Since impedance control is used only during the stance phase, we were prompted to consider
a different control strategy for the swing phase. These trajectories are tracked using
PD control. Thus, we proposed the cubic-Bezier-curve-based optimization to generate
appropriate trajectories for the given slopes during the swing phase, without any information
regarding the slopes. Before the heel contact occurs (terminal swing phase), low
gain PD control is used to adapt to the unexpected slopes and smoothly track the generated
trajectories. To validate the proposed framework, the concept was implemented on
a transfemoral prosthesis, AMPRO II, on various slopes. The main objective of the thesis
is to propose and verify a unified framework that enables the transfemoral prosthesis to
perform real-time inclined walking without a priori information regarding the terrain. | en |
