結合機器人/虛擬真實之智慧型人機介面

Abstract

為了加強在順應性工作執行時人類技巧的轉移, 我們在這篇論文中提出 一個智 慧型人機介面來結合機器臂與虛擬實境模擬系統. 這篇論文分為 兩部份, 在第 一部份, 我們提出了一個用於順應性工作的人類技巧轉移 方法. 一般說來, 在 處理順應性工作所碰到的困難主要導因於兩個理由. 一是很難清楚地去描述一 個順應性工作的執行過程. 二是即使執行相同 的工作, 但因為環境特徵的變化 而造成機器臂與工作環境產生不同的交 互作用. 有些學者就想要利用人類技巧 轉移的方法來解決上述的問題. 然而現今已提出的人類技巧轉移方法仍存在許 多不足, 因為人類與機器 臂在控制方法與機構上的不同. 所以我們在此提出一 個結合模糊邏輯及 基因法則架構的方法, 經由學習的能力來改善人類技巧的獲 取. 這個提 出的方法應用在去毛邊的順應性工作上來作為例證. 在第二部份, 我們 則發展一個結合機器人與虛擬真實的人機介面. 它不僅提供了一個可供模 擬的三維空間虛擬環境, 也提供了可與外部真實機器人通訊的介面. 操作 者可 藉由此介面同時控制虛擬及真實機器臂. 因為此介面可在真實實驗 前提供更多 的資訊, 當它被應用於電腦輔助設計及電腦輔助製造也是非 常有幫助的. In order to enhance human skill transfer for compliant task execution, in this thesis, we propose an intelligent human- machine interface to integrate robot manipulators with the virtual reality simulation system. The thesis is divided into two parts. In Part I, we propose a human skill transfer scheme for compliant tasks. Difficulties in dealing with compliant tasks mainly result fromtwo reasons. First, it is hard to describe the process of the compliant task execution precisely. Second, even for the same kind of compliant tasks, when theenvironment features vary, it may invoke different interaction between robot manipulators and the environment. Some researchers proposed using the technique of human skill transfer to tackle the problems aforementioned. The current humanskill transfer schemes are still with many deficiencies because the differences are present between the human and the robot manipulators in their mechanisms andcontrol strategies. We then propose a genetic-based fuzzy scheme to improve the ability in human skill acquisition via its learning capability. The proposed scheme is applied to the deburring compliant task for demonstration. In Part II,we develop a human-machine interface for Robot/VR integration. It not only provides a 3D virtual environment for simulations, but can also communicate withthe external robot manipulator. The operator can control the virtual and real manipulators via this robot/VR interface simultaneously. As the proposed interface can provide much information via simulations before experiment execution, it is also useful when applied to CAD (computer aided design) and CAM(computer aided manufacture).