A fuzzy system based on the obtained information of four infrared (IR) sensors and one electronic com-pass is proposed and implemented on a humanoid robot to avoid obstacles. A humanoid robot named TWNHR-3 with 26 degrees of freedom (DOFs) is de-signed so that it can do five basic motions. Four IR sensors and one electronic compass are installed on TWNHR-3 to detect the environment information including obstacles, the distances of the obstacles, and the directional angle of the robot. Based on the ob-tained information, an obstacle avoidance method is proposed to decide one behavior from five motions so that it can avoid obstacles and go to the destination area effectively. Some MATLAB simulation results with different number of obstacles and two real ex-periments are presented to illustrate the effectiveness of the proposed method.