The legged mechanism has flexibility to change its movement for ground conditions such as flat, slope, step or any other rough terrain. If mobile robots have such mobility and stability, they were able to work more wider fields and situations This paper proposes a method to acquire the gait pattern by using a genetic algorithm. A robot model which is based on real robot specifications is created to obtain the gait pattern through simulations. The objective of this proposed method is to achieve walking by applying the gait pattern to the real robot. The proposed method were evaluated by real robot experiments for loading the obtained gait pattern through simulations. There were difference between the real robot and the simulation models in the time. However, the robot was able to obtain the faster gait pattern and walking rear pattern by GA through simulations.