HGN Lab for Bionic Engineering
Mechanical Engineering | University of Utah

We developed a novel control architecture for robotic leg prostheses that adapts online at each step to provide a biological accurate function. The movement of the prosthesis during the swing phase adapts online to the specific user gait pattern using the minimum-jerk principle. Minimum-jerk provides an optimal way to program smooth movements, allowing amputees to obtain physiological gait symmetry at variable cadences, and potentially handling unexpected perturbations such as tripping.  Read More.

We are developing a novel control approach for powered prostheses that combines the neural commands from the user’s residual muscles with the mechanical sensors in the powered leg. This shared control approach provides users with the ability to voluntarily change the assistance provided by the powered prosthesis in a way that is both safe and intuitive. Ambulation tasks that benefits from this approach are standing up, squatting, and lunging.

Available prosthesis controllers use machine learning to understand the surrounding environment and adapt the prosthesis function. While functional in the laboratory, this approach struggles to deal with the variability typical of the real world. To address this issue, we are developing new methods to continuously adapt the prosthesis behavior to the user needs without an explicit classification of the environment. Using this approach powered prosthesis users can ambulate on any kind of stairs step-by-step, step-over-step, and even two steps at the time.

Adaptive Oscillators are a mathematical tool that can learn any periodic signal and provide a smooth estimate of its phase and offset. By using this innovative tool, we moved from a time-based to a phase-based control system for powered exoskeletons that is more robust to natural movement variability. Experiments show that this control approach can effectively provide assistance by reducing the user effort during walking. Read More.