基於機率之移動式抓取姿態規劃

Abstract

本論文研究使移動式機械臂在居家環境中規劃其對特定物體之抓取路徑的方法，使之移動至目標物時更能成功抓取。因為機器人的手臂有其工作範圍限制，當機器人前往目標物進行抓取時，若目標物在此範圍之外將導致抓取失敗。同時，機器人在環境中移動時產生的誤差同樣影響抓取的成敗。因此，為了提高抓取的成功率，本論文先計算出可抓取性地圖─藉由機率來描述機器人手臂在工作範圍內抓取的成功率；再計算出移動可達性地圖─機器人在環境中移動的難易度，同樣以機率表示。透過可抓取性與可達性地圖的結合，機器人可以決定一個移動最容易、抓取成功率最高的抓取位置。本論文以實驗室之雙臂機器人，對三種情境進行實驗，驗證機器人在此系統下更有效率地進行移動式抓取的任務。

In this thesis, a system is designed and experimented for efficient mobile manipulation of a dual-arm robot. Because of limited workspace of the robot arm, robot grasping may fail if the target object is outside the workspace. At the same time, navigation errors also affect the grasping as the robot needs to move toward the target object. In this thesis, we construct a graspability map to describe the possibility of successful grasping inside workspace of the robot arm and we also construct a reachability map to describe the feasibility the robot can navigate to the target. These two maps are integrated to determine the robot location for grasping the target object. Using a laboratory dual-arm robot, we conducted three experiments in different conditions to verify the efficiency of mobile manipulation task using the proposed method.