Modeling Human Walking with HZD-Based Control and Impulsive Toe-Off

1 Motivation Due to experimental difficulties, almost no scientific evidence to date definitively indicates that one lowerlimb prosthesis performs better than another [3]. A model of walking that is simple enough to allow systematic exploration of prosthesis design variables, yet detailed enough to accurately capture step dynamics, could help fill this gap. Unfortunately, such a model does not yet exist, even when the problem is simplified to a healthy subject. 2 State of the Art Human gait models range from highly dimensioned to very simplistic. Highly dimensioned models either mimic recorded human motion [6] or are too complex to easily use in iterative design [7]. Very simplistic models can provide information about a single part of gait, but not the whole gait [5]. Simple models controlled using hybrid zero dynamics (HZD) techniques [8] offer a promising compromise. Because they lack toe-off [2], though, current HZD-based models do not efficiently redirect their center of mass velocity like humans do [1]. As a result, HZD-based model gaits differ from human gaits to compensate for the lost energy. This research demonstrates that such simple models can be modified to have human-like energy transfer at impact. 3 Approach The symmetric, planar four-link model consists of a point mass at the hip (mass of the HAT), two massive thighs, and two massive shanks with attached circular feet [4] (Fig. 1). A step consists of a finite-time single support phase controlled using an HZD-based controller and an instantaneous impact to switch the stance leg. During impact, an additional impulse is applied at the hip to capture the effects of toeoff. Model gaits based on human walking data were created over a range of speeds using an optimization