Controlling Robot-Assisted Locomotion with Extended Kalman Filter Estimates of Phase and Activity

点击翻译按钮获取中文摘要

PROJECT ABSTRACT This diversity supplement to the parent R01 project will address our overall goal—to model and control human locomotion over continuously varying tasks—by indirectly estimating a gait-state comprising (1) the phase variable, (2) the phase rate, and (3) a selection of task variables such as the step length and ramp inclination. The phase variable is a signal that monotonically increases during the natural progression each human stride, and the phase rate is the rate of change of the phase variable. Task variables represent the physical characteristics of a locomotion task, such as the ground’s inclination or the speed of the human. Grouped together in the gait-state, these variables parametrize human leg kinematics and kinetics through a gait- mapping, which must be learned from gait data. The central hypothesis of this project is that continuously varying activities can be tracked in real-time by using the theory of state estimation to estimate this gait-state vector indirectly through a comparison between measured human kinematics and the kinematic predictions of the gait-mapping; this estimate can then be used to either imitate (in the prosthetic case) or augment (in the orthotic case) the human’s behavior. In this research, PhD candidate Roberto Manuel Leonardo Medrano III, BS (Leo) will be advised by PI Robert Gregg, PhD and Co-I Elliott Rouse, PhD, who will help him develop as a leader/researcher throughout the course of this project at the University of Michigan (U-M).