Augmenting Human-Inspired Phase Variables Using a Canonical Dynamical System

Accurately parameterizing human gait cycles is crucial for developing control systems for lower-limb wearable robots. Previous studies on characterizing human gait cycle investigated the use of human-inspired phase variables, but they lack robustness and adaptability to accommodate for transitions that occur between locomotor tasks. This paper proposes augmenting the traditional phase variables with a canonical dynamical system. This system considers the differences between the estimated and true thigh angular velocities, based on which the Fourier series coefficients and phase information will be dynamically updated via an adaptive frequency oscillator. Experiments with six able-bodied participants on level ground, inclines, and declines walking show that the augmented phase variable adapts to various locomotor tasks and transitions better than the traditional approach.