Metabolic cost breakdown of human walking: contributions from step frequency and length

The metabolic cost of walking is amenable to empirical curve fits, such as the increasing speed-related cost. A typical fit is derived from walking with preferred step parameters (self-selected step length and step frequency), but cannot generalize to other step parameter combinations. Mechanistic models may have more general, predictive ability. We therefore tested a simple dynamic walking model against empirical curve fitting. The model suggests mechanistic trends associated mainly with increasing step length, and with increasing step frequency. We used metabolic cost data for humans (N=10) walking at non-preferred step combinations to fit model parameters, which were then used to predict preferred walking. The model predicted the walking cost (R^2 = 0.59) and preferred step parameters (R^2 = 0.68) reasonably well. Mechanistic models may provide greater predictive value than pure curve fits.