Zero-Moment Point (ZMP) is the most popular concept that is applied to stabilize the gait motion of a biped robot. However, the computation of the ZMP position and the dynamic model is highly complex, and so requires a considerable time. Therefore, to eliminate complex computation, this work designs a new locomotion control system for humanoid robots, which yields walking gait characteristics that are similar to those of a human. Accordingly, the first goal of this study is to elucidate walking characteristics by measuring human motion during locomotion. It analyzes biped locomotion from the perspective of torque control at each joint. Because of the complexity of biped locomotion systems, a dynamic biped law cannot be realized by the direct application of modern control theory. Hence, fuzzy logic concept will be adopted in dynamic stabilization analysis and locomotion control. The fuzzy control is based on coordination control rules for dynamic stabilization that govern the actions of the biped robot link. The performance of the system is evaluated by making the biped walk on a horizontal plane surface and climbing up a flight of stairs. Experimental results are presented to confirm the effectiveness and applicability of the proposed fuzzy stabilization tuning approach.