Computational predictions of subject-specific muscle and knee joint contact forces during walking may improve individual rehabilitation treatment design. Such predictions depend directly on specified model parameter values. However, model parameters are difficult to measure non-invasively. Methods for muscle model parameter calibration have been developed previously. However, it is currently unknown how the musculoskeletal system chooses muscle model parameter values. Previous studies have hypothesized that muscles avoid injury during walking by generating little passive force and operating in the ascending region of the force-length curve. This hypothesis suggests that muscle model parameter values may be selected by the body to minimize passive force. The purpose of this study was to develop a method for calibrating muscle model parameter values and muscle moment arms during walking via minimization of passive force.