Fig. 2

Flow of data for estimating and evaluating within-stride metabolic cost. (A). A perturbed dataset was gathered using force perturbations at the COM. Biomechanical time series (e.g., kinematics, kinetics, muscle activations) as well as stride-mean metabolic cost were measured for each walking condition. (B). These measurements are stride normalized and (C). then converted to a stride-mean for each walking condition. (D). The stride means for each biomechanical measurement are custom standardized by subtracting the unperturbed stride mean from each perturbed stride mean and then dividing by the range of deviations from unperturbed walking. (E). The custom standardized biomechanical time series are then compared to the custom standardized within-stride metabolic cost using the sum of square error. This process will be iterative, where an additional custom standardized biomechanical time series may be added if it reduces the sum of square error. (F). The biomechanical time series or combination of biomechanical time series that corresponded to the lowest sum of square error are selected. The unperturbed condition from the selected biomechanical time series is used as the estimate for within-stride metabolic cost. (G). The original model-based within-stride metabolic cost is only used for validation of our perturbation-based method. Our perturbation-based method leverages information from stride-mean values that are experimentally available to indirect calorimetry