A design and implement method of a humanoid robot for skating is proposed in this thesis. The implemented humanoid robot not only can do basic skating motion, but also can make suitable decision to get up, balance, and navigation by itself based on the detected information from the accelerometer, force sensor, and CMOS sensor. A mechanism structure with 26 degrees of freedom (DOF) is design and four kinds of servo motors with different torque are used so that the implemented humanoid robot can do some basic skating motion. The digital signal process of μ’nsp made by SUNPLUS and the embedded soft core processor of NIOS II made by Altera are used to construct a two-processor control scheme. One accelerometer is installed on the robot to detect whether the robot falls down or not and a control strategy is made to let the robot get up by itself. Force sensors are installed on the feet of robot to calculate its Zero Moment Point (ZMP) and difference of ZMP. A two-input-and-one-output fuzzy balancing system is proposed to modify the angle of some servo motors so that the robot can also balance when it skates on an inclined plane. One CMOS sensor is used to capture the environment image and the image process is implemented on u’nsp to detect the lines on the floor. The strategy is design and implemented on u’nsp so that the vision-based robot can do some appropriate skating motion to follow the line. From the experiment results, we can see the implemented skating humanoid robot actually can do some basic skating motion, get up by itself, autonomous balance, and vision-based autonomous navigation.