Newborns with lower limb motor deficiencies do need rehabilitation. Soft exoskeletons based on soft artificial muscles are promising for this purpose. However, the highly non-linear dynamic behavior of their system makes very difficult to obtain an accurate model to be controlled. This paper proposes to use a free-model method called Incremental PI-Like Fuzzy logic controller (PI-Like FLC). This variation facilitates the design and interpretation of the fuzzy rules; furthermore, we were capable of using optimization algorithms and fuzzy-C-means to obtain the fuzzy sets and membership functions based on data obtained from a previously tuned PID controller. The PI-Like FLC was physically implemented on infant dummies of 0 and 6 months of age to control the knee flexion-extension motion, and a PID was used as a comparison. The results demonstrate a successful tracking control and exhibit robustness against parametric uncertainties and disturbances for both dummies' motion without the need of tuning the controller.