本論文設計一個模糊分散平衡控制器不僅可以使得小型人形機器人跨越障礙物，而且在遭受干擾時亦可以維持自我平衡地跨越障礙物。所建議的人形機器人包含頭部共有21個自由度，而其高度與重量分別為55公分與3.7公斤。在設計與實現上，主要有三大部分：動作設計、姿態擷取、以及平衡控制。在動作設計上，使用人機介面設計小型人形機器人跨越障礙物所需的標準動作。在姿態擷取上，在機器人身體內安裝兩個單軸的陀螺儀與一個具有三軸的加速度計來偵測身體姿態的響應，並使用SOPC (System on a Programmable Chip)技術平行擷取感測器的資料，接著以巴特沃斯濾波器與卡爾曼濾波器獲得較準確的機器人之Pitch和Roll之角度響應。在平衡控制上，首先經由人機介面獲得穩定跨越障礙物所得到的Pitch和Roll之角度響應，作為動態平衡控制所需的參考訊號，接著應用所設計的模糊分散平衡控制器使得人形機器人可以維持自我平衡地跨越障礙物。由實驗結果得知，所設計的模糊分散平衡控制器確實可以使得人形機器人於跨障礙動作中維持自我平衡，並且在受到外力干擾後亦可以恢復到平衡的狀態。 This thesis presents a fuzzy decentralized balance controller for a small size humanoid robot to step over the known obstacle and to keep dynamic balance in the face of external load. The proposed humanoid robot possesses 55 centimeter height, 3.7 kilogram weight, and 21 degree of freedom (DOF). This thesis contains the following three parts: motion control, posture capture and balance control. The first part designed a standard motion of stepping over an obstacle by human-machine interface. A set of sensors combining with two single-axis gyros in the pitching and rolling directions and one three-axis accelerometer, is employed to capture the posture of humanoid robot with respect to its center of gravity. Based on SOPC (System on a Programmable Chip), these signals are processed in the parallel type. Before applying a Kalman filter for the fusion of angular velocity and angular position, a Butterworth filter is applied to filter the unnecessary high frequency of the accelerometer. The final part of this thesis is the design of a fuzzy decentralized control to keep the dynamic balance of the humanoid robot during the stepping over an obstacle. Finally, the corresponding experiments of the stepping over an obstacle in the absence or presence of external disturbance confirm the effectiveness of the proposed humanoid robot system.
