- 學位論文
基於行為控制與規劃實現機器手臂操作任務
Implementation of robot manipulative tasks based on behavior control and plan
摘要
機器手臂應用多為重複性高、精準性高的工作，一旦操作環境改變，必需重新教導與規劃機器手臂動作，或者是在原本所規劃的動作裡，在其中插入新的動作，這樣過程是非常繁雜與耗時。在本研究，我們提出一個使用XABSL( Extensible Behavior Agent Specification Language)的機器手臂軟體編程平台，透過高階的語言的敘述規劃控制機器手臂與週邊感測與影像設備完成工作任務。因此本研究提供由影像辨識人類示範所得到的動作序列來規劃機器手臂的任務，也可以直接透過高階語言的編程控制機器手臂自行完成工作任務。在本研究中，三種工作任務來說明本機器手臂軟體編程平台的特色，(1)第一、透過影像辨識人類示範所得到的動作序列來進行機器手臂任務規劃。主要想法是從人類的連續動作辨識出人類當前的姿勢或手勢，並得到動作序列，序列中的動作一對一對應到本系統事先規劃好的動作單元，讓機器手臂透過動作單元自主地完成工作任務。(2)第二、透過影像辨識與魔術方塊求解演算法可得魔術方塊的轉動序列，解決魔術方塊為我們測試平台的實驗。(3)第三、使用三台機器手臂進行電源模組之組裝作業，以測試平台的應用性和拓展姓。實驗說明本研究方法經由人類示範或者是影像辨識魔術方塊，所辨識出來的工作任務動作序列可自主地由單手或雙手的機器手臂動作序列來完成。
並列摘要
In this study, we proposed a robotic arm programming platform to control robotic arm through a high-level language with imaging equipments to complete tasks. Three tasks were validated for the proposed robotic arm programming platform. (1)Applying image recognition to get human motion sequences from human demonstration and plan robotic manipulative tasks: The main idea is that human motion recognition from a continuous human gesture is used to derive the robot motion sequences. The robot motion sequences is constructed by robot motion blocks that are one-to-one corresponding mapping from elementary motions of the human demonstration. (2) Applying image recognition and Rubik's Cube solving algorithms to derive robot motion sequence. In this study, used XABSL (Extensible Behavior Agent Specification Language) as the key component of the proposed robotic arm programming platform. (3) Using three robot arms in an assembly task to demonstrate the feasibility of the proposed programming framework. The experiments were used to validate the platform which could autonomously complete the human demonstration task or solve the Rubik's Cube task.