Individuals with stroke often have diminished muscle strength that contributes to impaired push-off propulsion during walking, resulting in abnormal compensatory mechanisms and slower gait speeds. In this paper, we present a sagittal, 2-link biomechanical model of ankle push-off propulsion dynamics. The purpose of the model is to predict the parameters of an impedance controller to generate a desired level of anterior-posterior push-off during assisted walking with a modular ankle robot (“Anklebot”). In conjunction with our development of a novel gait event-triggered training approach, this model will facilitate tailoring the assisted push-off to individual severity. Here, we develop the model from first principles and experimentally validate it in a healthy subject.