PhD Porjects:
Step Controller Identification in Perturbed Walking
Foot placement controller were identified from nine participants’ walking data
at three walking speed (0.8, 1.2, 1.6 m/s). Results showed that capture theory
is not a bad estimation of how humans choose their foot placements, but a
little bit conservative. In addition, control gains vary with walking speed,
instead of constant as previous understanding.
Virtual Cycling Machine using Indego
An Indego exoskeleton was used as a force feedback device to provides corresponding
resistance torques at hip & knee joints for users to do cycling. Dynamic
parameters and friction of the Indego was identified. Robust impedance
control turned the Indego into a mass-spring-damper system.
Impedance Controller Identification in Human Walking unber Perturbation
Phase dependent impedance controllers are identifying from nine participants’
walking data at three walking speed (0.8, 1.2, 1.6 m/s). We are trying to
show that a single perturbation (treadmill belt speed) is sufficient to identify
parametric impedance controllers in all leg joints (hip, knee, and ankle). With
this, we will be able to identify impedance parameters of hip and knee joints
from regulate perturbed data.
Perturbed Walking Experiment
Collaborating with Farzad Ehtemam, 13 Gigabyte perturbed walking data
was collected. Twenty-five young adults participated this study. Perturbations
are in both AP and ML directions. Recorded information includes kinematics
(motion capture marker data), ground reaction force (2x6 dofs), and muscle
activations (11 EMG sensors).
Controller Identification in Human Standing Balacne Task
Standing balance experiment was conducted with random square wave perturbation.
Multiple types controllers were identified from the experiment data. Results
suggested that control system with cross joints feedback can better explain the
experimental data. Nonlinear and time delay property inside control system
helps improve the explanation also.
Ball Bouncing Optimization
The optimization speed of Ipopt and Snopt was examined through the ball
bouncing trajectory optimization. Ball bouncing is the simplest system which
has the similar strong nonlinear property as the landing in human walk.
Results showed that Ipopt (ma-87) is slightly faster than Snopt.