This paper presents a novel Gait Pattern Generator (GPG) developed for the \Alice" social humanoid robot, which up to now lacked an appropriate walking pattern. Due to the limitations of this robot, the proposed gate pattern generator was formulated based on a nine-mass model to decrease the modeling errors and the inverse kinematics of the whole lower-body was solved in such a way that the robot remained statically stable during the movements. The main challenge of this work was to solve the inverse kinematics of a 7-link chain with 12 degrees of freedom. For this purpose, a new graphical-numerical technique has been provided using the de nition of the kinematic equations of the robot joints' Cartesian coordinates. This method resulted in a signi cant increase in the solution rate of calculations. Finally, a novel algorithm was developed for step-by-step displacement of the robot towards a desired destination in a two-dimensional space. Performance of the proposed gate pattern generator was evaluated both with a model of the robot in a MATLAB Simulink environment and in real experiments with the Alice humanoid robot.