傳統上常使用預設的ZMP值來求解出二足機器人軀幹所需補償軌跡，本文不使用預設ZMP值，而設計模糊邏輯ZMP產生器求出ZMP值，其方法是經由所量測到腳掌和腰部幾何中心的位置，利用模糊算術的方法，求解出能使機器人穩定步行的ZMP值，且進一步設計一個八連桿二足機器人，利用MSC. VISUAL NASTRAN 4D與MATLAB Fuzzy Tool Box、MATLAB SIMULINK結合來模擬，證實模糊ZMP產生器能提供機器人穩定的步行設計，並且使用Verilog HDL 設計模糊ZMP產生器的IC，並利用ModelSim模擬，並驗證其功能的正確性。 Traditionally, it is often to use preset ZMP to solve the necessary compensation trajectory of biped robot trunk. This thesis does not use preset ZMP, instead, a fuzzy logic ZMP generator was designed to estimate the ZMP values. This method measures the position of geometrical center of sole and waist, then utilizes a set of fuzzy arithmetic equations to calculate the ZMP values.The robot design based on those ZMP values can make the robot walk steadily. In addition, a eight link biped robot was designed. The simulation was completed by combining several tools such as MSC., VISUAL NASTRAN 4D, MATLAB Fuzzy Tool Box and MATLAB SIMULINK. The simulation proved that fuzzy ZMP generator can provide a stable biped robot design. Verilog HDLwas used to design the Fuzzy ZMP Generator IC,and ModelSim was used to simulate and verify the desired functions of ZMP generator.