本論文提出一個基於力道預測和貪婪演算法之可形變物件裝配的方法。其結合了硬體技術、影像處理技巧和計算方法，並且整個系統是使用可以快速搭建機器人軟體框架之機器人操作系統(Robot Operating System, ROS)來建立的。本論文分為三個主要部分：（a）力控制和視覺輔助，（b）線性迴歸和貪婪演算法的整合應用，以及（c）實體機械手臂的實驗方法。本論文使用一個具有力/力矩感測器之六個自由度的機械手臂UR5e，並且透過力控制和力矩控制來控制機械手臂的三維運動。本論文使用Realsense D435i相機進行機械手臂之視覺輔助。將相機所擷取之皮帶輪的裝配點進行座標轉換後的座標位置傳送給系統去控制機械手臂。本論文提出一個基於力道資料集線性回歸模型的貪婪演算法，能夠預測其在不同位置上所受到的力，並根據這些預測結果來確定機械手臂的裝配路徑。本論文使用視覺辨識來簡化公式計算的難度、整理機械手臂之運動規劃與力學公式、分析可形變物件沿著整個路徑的受力情形、和使用貪婪演算法來求取裝配路徑點位。所提方法能夠優化總受力和路徑長度，並讓系統能夠在三維空間中快速找到合適的路徑。最後，一些模擬以及實驗結果的說明，所提的系統確實能夠有效地讓機械手臂完成可變形物件的裝配。

This thesis proposes a deformable object assembly method based on force prediction and greedy algorithm. It combines hardware technologies, image processing techniques, and computational methods, and the entire system is built using the Robot Operating System (ROS) that can quickly build a robot software framework. The thesis is divided into three main parts: (a) force control and visual assistance, (b) integrated application of linear regression and greedy algorithm, and (c) experimental method of physical robot manipulator. A six-degree-of-freedom robot manipulator UR5e with a force/torque sensor is used, and the three-dimensional movement of the robot manipulator is controlled through force control and torque control. The Realsense D435i camera is used for visual assistance of the robot manipulator. The coordinate position after coordinate conversion of the assembly points of the pulley captured by the camera is transmitted to the system to control the robot manipulator. This thesis proposes a greedy algorithm based on linear regression model of the force data set, which can predict the forces it receives at different positions, and determines the assembly path of the robot manipulator based on these prediction results. This thesis uses visual recognition to simplify the difficulty of formula calculation, organizes the motion planning and mechanical formulas of the robot manipulator, analyzes the force conditions on deformable objects along the entire path, and uses a greedy algorithm to obtain assembly path points. The proposed method can optimize the total force and path length, and allow the system to quickly find a suitable path in three-dimensional space. Finally, some simulation and experimental results show that the proposed system can indeed effectively enable the robot manipulator to complete the assembly of deformable objects.

中文摘要	I
英文摘要	II
目錄	III
圖目錄	V
表目錄	VII
第一章 緒論	1
1.1研究背景	1
1.2研究動機與目的	8
1.3論文架構	10
第二章 系統架構與硬體設備	12
2.1硬體設備之規格	12
2.2開源軟體	16
2.3系統架構	20
第三章 力道控制與影像輔助	22
3.1力道資料的預處理	22
3.2 影像視覺輔助	24
3.3 機械手臂運動學控制	32
第四章 線性迴歸與貪婪演算法整合運用	35
4.1線性迴歸力道模型的使用	35
4.2貪婪演算法演算法	36
4.3 機械手臂演算法應用整合	41
第五章 實驗結果	45
5.1 力道預測模型與比較	45
5.2貪婪演算法路徑規劃	47
5.3真實機械手臂之裝配任務執行	54
第六章 結論與未來展望	56
6.1 結論	56
6.2 未來展望	57
參考文獻	59
附錄1：符號對照表	63
附錄2：中英文對照表	65

圖1.1、力道數據化的應用與驗證[6]	3
圖1.2、根據力道感測實體應用[7]	4
圖1.3、力道控制預測捕捉圖[9]	5
圖1.4、可形變物件影像處理成果圖[10]	6
圖1.5、可形變物件裝配流程及力道感測數據圖[13]	7
圖1.6、線路徑規劃示意圖 [17]	8
圖1.7、2021 WRS 可形變物件裝配任務[19]	9
圖1.8、工業自動化可形變物件裝配問題[22]	10
圖2.1、裝配任務實際實驗環境	13
圖2.2、機械手臂UR5e的實體圖	13
圖2.3、HIWIN夾爪與機構之實體圖	13
圖2.4、Intel® RealSense D435i 深度攝影機之實體圖	14
圖2.5、ROS通訊架構圖	16
圖2.6、ROS系統架構圖	18
圖2.7、可形變物件裝配架構圖	20
圖3.1、線性迴歸具體示意圖展現[25]	23
圖3.2、座標轉換示意圖[26]	25
圖3.3、(u, v)座標軸示意圖[27]	26
圖3.4、RViz中tf之座標系示意圖[28]	28
圖3.5、相機與物體遠近影響示意圖[28]	29
圖3.6、手眼校正示意圖[29]	30
圖3.7、D-H連桿參數示意圖[30]	33
圖4.1、貪婪演算法結合多目標點運動規劃方法[33]	37
圖4.2、可形變物件裝配系統流程圖	42
圖4.3、貪婪演算法系統架構圖	43
圖5.1、原始數據與力道預測對照圖	47
圖5.2、力道和剩餘距離沿路徑的比較結果圖	49
圖5.3、正規化累積力道沿路徑的比較結果圖	51
圖5.4、3D路徑比較結果圖	53
圖5.5、影像輔助處理示意圖	54
圖5.6、裝配皮帶與皮帶輪物件圖	55
圖5.7、實際機械手臂可形變物件裝配分鏡圖	55
圖6.1、強化學習環境示意圖	58
圖6.2、多機器人協作裝配環境[35]	58

表 2.1、個人電腦之軟硬體規格表	13
表 2.2、UR5e之硬體規格表	14
表 2.3、UR5e之力量/力矩感測器規格表	14
表 2.4、HIWIN XEG-32-C23L3-W2-S夾具之規格表	15
表 2.5、Intel® RealSense D435i 深度攝影機之規格表	15
表 2.6、ROS節點彙整表	19
表 3.1、D-H連桿參數定義[30]	33
表 5.1、三種不同實驗距離的資訊表	48

[1]	Zhu, Jihong, et al. "Challenges and Outlook in Robotic Manipulation of Deformable Objects," IEEE Robotics and Automation Magazine, vol. 29, no. 3, 2022, pp. 67-77. doi: 10.1109/MRA.2022.3147415.
[2]	Gu, Feida, et al. "A Survey on Robotic Manipulation of Deformable Objects: Recent Advances, Open Challenges and New Frontiers," arXiv preprint arXiv:2312.10419.
[3]	Salzmann, T., Kaufmann, E., Arrizabalaga, J., Pavone, M., Scaramuzza, D., Ryll, M. "Real-time Neural MPC: Deep Learning Model Predictive Control for Quadrotors and Agile Robotic Platforms," IEEE Robotics and Automation Letters, vol. 8, no. 2, 2023, pp. 2341-2348. doi: 10.1109/LRA.2023.3246839.
[4]	Johannsmeier, L., Gerchow, M., Haddadin, S. "Towards Autonomous Robotic Assembly: Using Combined Visual and Tactile Sensing for Adaptive Task Execution," Journal of Intelligent & Robotic Systems, 2020, pp. 34-45.
[5]	Zhang, M., Hauser, K. "Semi-Infinite Programming with Complementarity Constraints for Pose Optimization with Pervasive Contact," Proceedings of the 2021 IEEE International Conference on Robotics and Automation (ICRA), 2021, pp. 2345-2352. doi: 10.1109/ICRA48506.2021.9561987.
[6]	Zhu, Jihong, et al. "Dual-arm robotic manipulation of flexible cables," IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2018, pp. 1234-1240.
[7]	Sun, Zhanyi, et al. "Force-Constrained Visual Policy: Safe Robot-Assisted Dressing via Multi-Modal Sensing," IEEE Robotics and Automation Letters, vol. 9, no. 5, 2024, pp. 8910-8917.
[8]	Wang, Dongxiao, Xing Wu, Kai Xu, and Xiaoqin Liu. "Research on Gravity Compensation of Robot Arm Based on Model Learning," IEEE Robotics and Automation Letters, vol. 8, no. 5, 2023, pp. 1234-1241.
[9]	Navarro-Alarcon, D., Liu, Y.-H., Romero, J. G., Li, P. "Model-Free Visually Servoed Deformation Control of Elastic Objects by Robot Manipulators," IEEE Transactions on Robotics, vol. 32, no. 5, 2016, pp. 1233-1243. doi: 10.1109/TRO.2016.2608885.
[10]	Tang, T., Wang, C., Tomizuka, M. "A Framework for Manipulating Deformable Linear Objects by Coherent Point Drift," IEEE Robotics and Automation Letters, vol. 3, no. 4, 2018, pp. 3426-3433. doi: 10.1109/LRA.2018.2852770.
[11]	Fletcher, R., Leyffer, S. "Solving Mathematical Programs with Complementarity Constraints as Nonlinear Programs." Optimization Methods and Software, vol. 19, no. 1, 2004, pp. 15-40. doi: 10.1080/10556780410001654241.
[12]	Schwartz, A. "Mathematical Programs with Complementarity Constraints: Theory, Methods, and Applications," PhD Dissertation, Julius-Maximilians-Universität Würzburg, 2011.
[13]	Jin, Shiyu, et al. "Trajectory Optimization for Manipulation of Deformable Objects: Assembly of Belt Drive Units," arXiv preprint arXiv:2106.00898 (2021).
[14]	Pecyna, Leszek, Siyuan Dong, and Shan Luo. "Visual-Tactile Multimodality for Following Deformable Linear Objects Using Reinforcement Learning," IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2022, pp. 2215-2222.
[15]	Qin, Yili, Adrien Escande, Fumio Kanehiro, and Eiichi Yoshida. "Dual-Arm Mobile Manipulation Planning of a Long Deformable Object in Industrial Installation," IEEE Robotics and Automation Letters, vol. 8, no. 5, 2023, pp. 9956-9963.
[16]	Liu, J., Fan, P. "Improved RRT-connect algorithm based robotic arm path planning," Computer Engineering and Applications, 2021, pp. 45-52.
[17]	Keipour, Azarakhsh, Maryam Bandari, and Stefan Schaal. "Efficient Spatial Representation and Routing of Deformable One-Dimensional Objects for Manipulation," IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2022, pp. 2134-2141.
[18]	Sidki, Nahid. “Autonomous Pick and Place for Mechani-cal Assembly WRS 2020 Assembly Challenge,” Journal of Robotics and Mechatronics, vol. 35, no. 1, 2023.
[19]	World Robot Summit. "Assembly Challenge | Industrial Robotics Category | World Robot Challenge," World Robot Summit, https://wrs.nedo.go.jp/en/wrs2020/challenge/industrial/assembly.html.
[20]	Caporali, A., Galassi, K., Palli, G. "Deformable Linear Objects 3D Shape Estimation and Tracking From Multiple 2D Views," IEEE Robotics and Automation Letters, vol. 5, no. 3, 2020, pp. 3796-3803. doi: 10.1109/LRA.2020.2976904.
[21]	Moll, Mark, and Lydia E. Kavraki. "Path Planning for Deformable Linear Objects," IEEE Transactions on Robotics, vol. 22, no. 4, August 2006, pp. 625-636. doi: 10.1109/TRO.2006.878933.
[22]	Kimble, K., Albrecht, J., Zimmerman, M., Falco, J. "Performance Measures to Benchmark the Grasping, Manipulation, and Assembly of Deformable Objects Typical to Manufacturing Applications," Frontiers in Robotics and AI, vol. 9, 2022. doi: 10.3389/frobt.2022.999348.
[23]	MoveIt, URL: https://fugjo16.blogspot.com/2017/05/ros-moveit.html
[24]	Habib, A., Bhakta, R. "Path Planning for Autonomous Vehicles Using Model Predictive Control," University of California, Santa Cruz, 2023.
[25]	Harbulot, Julien. "Polynomial basis expansion," PhD Thesis, University of Lyon, 2018. 
[26]	openMVG_libraries_cameras,URL:https://openmvg.readthedocs.io/en/latest/openMVG/cameras/cameras/.
[27]	針孔成像，URL:https://silverwind1982.pixnet.net/blog/post/134551091.
[28]	歐燦坤，基於近端策略最佳化之線上3D裝箱，淡江大學電機工程學系碩士論文(指導教授：翁慶昌)，2023.
[29]	蔡岱庭，應用於機械手臂之通用型自動手眼校正系統，淡江大學電機工程學系碩士論文(指導教授：翁慶昌)，2023.
[30]	楊奕禾，基於深度學習之模組化機械手臂的設計，淡江大學電機工程學系博士論文(指導教授：翁慶昌)，2023.
[31]	Li, H., Duan, J., Qin, J., Zhou, Y. "Multi-Objective Point Motion Planning for Assembly Robotic Arm Based on IPQ-RRT* Connect Algorithm," Actuators, vol. 12, no. 12, 2023, Article 459. doi: 10.3390/act12120459 (MDPI).
[32]	Cui, C., Chen, S., Shen, C., Li, F. "An improved bi-RRT-based path planning algorithm for mobile robots," Computer Measurement & Control, vol. 30, no. 5, 2022, pp. 123-130.
[33]	Zhang, J., Wang, H., Guo, Y., Yang, F., Zhao, Q. "Research on Dual-Arm Robot Assembly Path Planning Based on Improved RRT Algorithm," Proceedings of 2022 Chinese Intelligent Systems Conference. Lecture Notes in Electrical Engineering, vol. 951, 2022, pp. 451-458. Springer, Singapore. doi: 10.1007/978-981-19-6226-4_75