Modular 7 DoF Dual Arm Robot with 3D Visual Perception and Manipulation

使人類成為萬物之靈，與其他動物有所區別的關鍵就在於人類有靈 巧的雙手。因此在機器人領域中，針對類人型雙手臂機器人的動作規 劃研究，也就成為一個重要的議題。本篇論文重點為，發展具有視覺 回授功能、冗餘自由度的模組化七自由度雙手臂機器人，推導其運動 學的解析解，並利用視覺感知功能完成辨識工作桌面物體位置，使動 作規劃器可規劃出夾取環境物體，並適當的運用雙手的優勢－在不互 相碰撞的情況下，可同時夾取兩個物體。並同時透過模擬與實驗測試 該動作規劃器的效果，以使本雙手臂機器人也能做出如同人類般靈巧 的動作。 本系統以兩支七自由度模組化機械手臂作為研究對象，以架構在控 制器區域網路通訊協定 (Controller Area Network, CAN) 的 CANopen 作 為通訊協定。最後加上具有實時作業系統－以確保運算時脈的穩定性 －做為控制命令、動作規劃演算法的運算器。 為具有七自由度的手臂，安裝馬達的部分設計成模組化，可大幅減 少設計不同部位零件所花的時間、增加不同部位零件的可替換性、減 少安裝、配線、維修的複雜度。 傳統上在計算順向、逆向運動學時，使用雅可比 (Jacobian) 矩陣做 關節空間和卡式空間下的解的轉換。然而此兩者間的關係是非線性的， 使用雅可比矩陣會造成解的誤差。並且，由於以雅可比矩陣計算逆運 動學時，需針對多維矩陣運算，故也較使用解析解費時。故推導具七 自由度冗餘關節的手臂的運動學解析解，可使本系統在做順向、逆向 運動學時，能精確、迅速求出關節空間或卡式空間下的解。 最後，結合一具三維環境感測功能的感測器，應用在本雙手臂機器 人上，使雙手臂機器人可完成如人類一般，辨識目標物體位置後夾取 該物體的能力。若桌上有多個物體時，本機器人也具備任務規劃的能 力，以使本機器人可利用存在雙手的優勢，以較快的速度完成夾取所 有物體的任務。Having dexterous arms makes human different from other creatures. In the field of robotics, research on anthropomorphic becomes an important issue. In this work, we proposed a modularized 7-DoF dual arm robot with 3D visual perception ability. First, we derived the analytical inverse kinematics solution for both arms. Second, visual sensing is also implemented, such that the robot can recognize objects in the environment. Third, a planning algorithm is implemented to grant the robot the ability to plan the order to grasp all objects put on the working table. We use CANopen as the communication protocol. With real time operating system to make the timer stable, the control command and motion planning algorithm become more stable. Both manipulators in this work are consisted of modularized components, which makes the assembly, wiring, and maintenance faster and more efficient. In addition, it reduces the time spent on designing different parts of the robot, thus makes the maintenance easier. Last but not least, with modularized component, the configuration of the robot can be changed easily. When dealing with kinematics problem, traditionally numerical method was used to calculate inverse kinematics problem. However, the relationship between Cartesian space and joint space is not linear, using this method, i.e. using Jacobian matrix will cause error. Hence, we proposed an analytical inverse kinematics solution for a new configuration of 7-DoF manipulator of the modularized dual arm robot. With the solution, inverse kinematics problem can be solved faster and more accurate. At last, with 3D environment sensing ability, making the robot be able to recognize the position of target objects then grasp it like human. Also, when there are several objects put on the working table, the robot can generate the order to grasp objects using the advantage of possessing dual arm, thus makes the grasping task finished faster.Table of Contents 誌謝 iii 中文ᄔ要 v Abstract vii Contents 1 List of Figures 3 List of Tables 7 1 Introduction 9 1.1 Background and Motivation . . . . . . . . . . . . . . . . . . . . . . . . 9 1.2 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 1.2.1 State-of-the-art Dual Arm Robot . . . . . . . . . . . . . . . . . . 10 1.2.2 Analytical Inverse Kinematics Solution . . . . . . . . . . . . . . 11 1.3 Thesis Organization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 2 Modularized Dual-Arm Robot 15 2.1 Mechanical Design of Robot Arms . . . . . . . . . . . . . . . . . . . . . 15 2.2 System structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 3 Kinematics Analysis 31 3.1 Robot Coordinate Configuration . . . . . . . . . . . . . . . . . . . . . . 31 3.2 Foward Kinematics Analysis . . . . . . . . . . . . . . . . . . . . . . . . 31 3.3 Inverse Kinematics Analysis . . . . . . . . . . . . . . . . . . . . . . . . 35 3.3.1 analytical inverse kinematic solution for the left arm . . . . . . . 37 3.3.2 analytical inverse kinematic solution for the right arm . . . . . . 44 3.3.3 singularaty issue . . . . . . . . . . . . . . . . . . . . . . . . . . 47 13.3.4 verification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 3.3.5 workspace . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 4 3D Perception 55 4.1 Object Capture System . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 5 Motion Planning 63 5.1 Object grasping planning . . . . . . . . . . . . . . . . . . . . . . . . . . 63 5.1.1 Control structure . . . . . . . . . . . . . . . . . . . . . . . . . . 63 5.1.2 Homing process . . . . . . . . . . . . . . . . . . . . . . . . . . 64 5.1.3 Grasping task planning . . . . . . . . . . . . . . . . . . . . . . . 65 5.1.4 Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68 6 Conclusions and Contributions 73 7 Future Works 75 References 77 VITA 8