Previous work demonstrated that adults naturally adopt a walking frequency to optimize physiological cost, symmetry, and stability. Furthermore, the optimal frequency is predictable using the force-driven harmonic oscillator (FDHO) model. However, no studies have established the developmental processes of optimization in children. Thus, the purposes of this study were to examine the predictability of the preferred stride frequency (PSF) and optimization features of 3- to 12-yr-old children using the FDHO model. Forty-five children and nine adults were measured for anthropometric data to calculate the predicted frequency. They later walked at three frequencies (PSF, PSF +25%, and PSF -25%) at a constant speed on a treadmill. The results indicated that the FDHO model was accurate in predicting the preferred frequency of children (prediction error < 0.07 s). We identified three stages of learning in the development of optimization: an early manifestation of sensitivity to resonant frequency, the subsequent development of ability to modulate walking frequency, and the final establishment of an adult optimization form at age seven. Our findings suggest that walking development may be determined by the dynamic cooperation of physiological, neural, and musculoskeletal systems with respect to the environmental context.