臺灣博碩士論文加值系統

隨著機器人的應用日益廣泛，生產線上動態取放的應用也逐漸增加，透過機器人視覺應用以及引入人工智慧，可以大幅降低傳統機械手臂所需要的輔助人力。本論文主要研究目的為機械手臂插銷控制應用，結合機器人學以及人工學習的機器人系統達成兩個任務：隨機取放、模擬人類手腕調整歪斜螺絲。
本文首先，實現在不同尺寸、角度以及明暗度的模板比對演算法，並設計簡易整合介面，達成隨機取放。並引入強化學習，結合遠端運動中心控制，教導機器人類比人類手腕轉動調整螺絲歪斜狀況。

Owing to the applications of robot arm become increasingly extensive, dynamic pick-and-place on the production line has gradually increased. Through robot vision system and artificial intelligence , the requirements of operation personnel can be greatly reduced. The purpose of this thesis is the robot arm plugging control application. Combining robotics and artificial intelligence into robot systems, two tasks are achieved: random pick and place, and simulation of human wrist adjustment skew screws.
In this paper, we first implement the rotation, scale, translation-invariant template matching algorithm , and design a simple integration interface to achieve random pick and place. Using reinforcement learning technique, combined with remote center of motion control, training the robot to learn how to adjust the skew of the screw like the human wrist did.

口試委員會審定書 #
誌謝 i
中文摘要 ii
ABSTRACT iii
CONTENTS iv
LIST OF FIGURES vii
LIST OF TABLES x
Chapter 1 緒論 1
1.1 文獻回顧 1
1.2 研究動機 3
Chapter 2 機械手臂分析 6
2.1 機械手臂運動學 6
2.1.1 位置姿態描述 7
2.1.2 齊次轉換 8
2.1.3 正向運動學 9
2.1.4 逆向運動學 12
2.2 關節空間與任務空間 12
2.2.1 賈可比矩陣 12
2.2.2 奇異點分析 14
Chapter 3 機械手臂控制理論 15
3.1 零空間控制 15
3.1.1 零空間矩陣 15
3.1.2 零空間應用於機械手臂控制 17
3.2 遠端運動中心控制 17
3.2.1 遠端運動中心 17
3.2.2 遠端運動中心矩陣應用於機械手臂控制 18
Chapter 4 強化學習 20
4.1 馬可夫決策過程 20
4.2 價值疊代與策略疊代 22
Chapter 5 研究方法與架構 27
5.1 視覺系統演算法設計 27
5.1.1 Circular sampling filter 28
5.1.2 Radial sampling filter 30
5.1.3 Template matching filter 32
5.2 視覺系統硬體 32
5.3 介面設計與通訊架設 33
5.4 強化學習參數設計 39
5.4.1 狀態空間 39
5.4.2 動作空間 43
5.4.3 回饋 44
5.4.4 價值疊代 45
5.5 機械手臂系統建模 47
5.5.1 正向運動學 47
5.5.2 逆向運動學 50
5.5.3 奇異點分析 53
5.6 機械手臂控制架構設計 54
5.6.1 零空間 54
5.6.2 遠端運動中心 55
Chapter 6 模擬與實驗結果 57
6.1 視覺系統實驗結果 57
6.1.1 螺絲姿態判斷 57
6.1.2 視覺系統實際夾取 58
6.2 強化學習模擬結果 61
6.2.1 零空間控制 61
6.2.2 遠端運動中心控制 67
Chapter 7 結論和未來展望 75
7.1 結論 75
7.2 未來展望 75
參考文獻 76

1.曲建仲, 機器是如何學習與進步？人工智慧的核心技術與未來. 科學月刊, 2019.
2.Wagner, S., Reinforcement Learning and Supervised Learning: A brief comparison. 2018.
3.Horn, B., Robot Vision. 1986.
4.Kumar, R., et al. Object detection and recognition for a pick and place Robot. in Asia-Pacific World Congress on Computer Science and Engineering. 2014.
5.wikipedia, Template matching.
6.Araújo, S. and H. Kim, Ciratefi: An RST-invariant template matching with extension to color images. Vol. 18. 2011. 75-90.
7.Silver, D., et al., Mastering the game of Go with deep neural networks and tree search. Nature, 2016. 529: p. 484.
8.Kober, J. and J. Peters, Reinforcement Learning in Robotics: A Survey, in Learning Motor Skills: From Algorithms to Robot Experiments, J. Kober and J. Peters, Editors. 2014, Springer International Publishing: Cham. p. 9-67.
9.Pane, Y.P., et al., Reinforcement learning based compensation methods for robot manipulators. Engineering Applications of Artificial Intelligence, 2019. 78: p. 236-247.
10.Nagendra, S., et al. Comparison of reinforcement learning algorithms applied to the cart-pole problem. in 2017 International Conference on Advances in Computing, Communications and Informatics (ICACCI). 2017.
11.Kober, J. and J. Peters, Policy search for motor primitives in robotics. Machine Learning, 2011. 84(1): p. 171-203.
12.Craig, J.J., Introduction to Robotics: Mechanics and Control. 1989: Addison-Wesley Longman Publishing Co., Inc. 450.
13.Yoshida, S., T. Kanno, and K. Kawashima, Surgical Robot With Variable Remote Center of Motion Mechanism Using Flexible Structure. Journal of Mechanisms and Robotics, 2018. 10(3): p. 031011-031011-8.
14.Aghakhani, N., et al. Task control with remote center of motion constraint for minimally invasive robotic surgery. in 2013 IEEE International Conference on Robotics and Automation. 2013.
15.Corke, P., Robotics, Vision and Control: Fundamental Algorithms in MATLAB. 2013: Springer Publishing Company, Incorporated. 594.