多指機器手掌抓取物體的分析,一直是機器人研究中十分重要的研究議題。日常生活中,物體並不會是個簡單的幾何形狀,例如方形、圓形、圓柱、錐形。而是像馬克杯、小花瓶、清潔劑等,較為複雜的形狀。而多指機器手掌的自由度比起機器夾爪多了許多,找到有效的抓取點就變得十分重要。因此,本論文旨在探討如何在複雜幾何的物體上利用力旋量空間尋找多指機器手掌可抓取的姿態以及根據物體邊界盒限制機器手臂可到達的區域加速找到一組可以抓取的最終姿態。再以阻尼最小平方法解決逆運動學中奇異點問題,並且在六軸機器手臂軸空間中,利用兩棵快速擴展隨機樹連接法尋找避開障礙物的路徑。並透過三維點雲進行點雲處理並重建環境以及物體表面模型,以達到自動化抓取的目標。
Analysis of a multi-fingered robot hand for grasping objects is an important research issue in robotics. Objects in daily life are not simple geometric shape like square, spherical, cylindrical, conical shapes. Most are the complex geometric shapes. Since the number of degrees of freedom of a multi-fingered robot hand is more than the simple gripper, it is important to find the grasping points. This thesis mainly focuses on how to determine the grasping points on the complex geometric shapes with the multi-fingered robot hand. Each grasping can be described by grasp wrench space (GWS). The quality measure can be determined from analyzing the GWS. The bounding box of object is a constraint on the workspace of the robot hand-arm system. According to the constraints, the random based algorithm can speed up the searching speed. The damped least square (DLS) method is use for solving the inverse kinematics problem about singular points. The rapidly-exploring random trees (RRT) – connect algorithm is used to search for the collision-free path in the joint space of the robot arm. The point cloud can be processed by point cloud processing and surface reconstruction. It combines the real environment and simulation. The goal can be reached through these methods.