移動式機械臂之抓物控制設計

Abstract

本論文的目的在設計一移動式機械臂的抓物控制系統，藉由機械臂上攝影機之影像資訊，迴授控制移動式機器人本體與機械臂的速度，以達成自動抓取的功能。本文基於Image Jacobian及機器人與攝影機間的Jacobian關係，設計出一個基於影像的視覺伺服控制器。此控制器可同時產生機器人本體與手臂的速度命令以追蹤空間中之物體特徵。在影像特徵擷取方面，本文利用一套嵌入式影像處理平台，採用Eye-in-hand的攝影機架設方式，以Pattern matching的方法尋找目標物體的角點以作為伺服控制器的輸入影像特徵點。文中以目標影像中兩種不同的角點作為視覺伺服控制器的輸入，設計機器人接近目標物體的行為。在控制設計上採用Behavior-based 控制架構來結合兩種接近行為與一個抓取行為，可引導移動式機械臂到達目標物所在地點，進而以單自由度之手爪抓取目標物。透過實驗驗證，此設計可導引一部擁有4自由度的移動式機械臂成功抓取特定之目標物。

The objective of this study is to design a grasping controller for a mobile manipulator. By controlling the velocity of the mobile robot and the 2-DOF robot arm simultaneously, the grasping of an object is realized. A visual servo controller has been realized by adopting the concept of image Jacobian, and the Jacobian of mobile manipulator and camera. This controller determines the robot body and robot arm’s velocity commands. We define the corners of the target image as the feature points which will be the input of visual servo controller. An embedded vision system is utilized to extract these feature points by pattern matching method. With two kinds of feature points on the target image and the visual servo controller, we design two approaching behaviors which will be activated when target is near or far away from camera, respectively. Combining two approaching behaviors and a grasping behavior, we use a behavior-based control architecture to realize an autonomous grasping behavior of a mobile manipulator. With this design, the 4-DOF mobile manipulator can find and grasp a target successfully.