可變與自順應外形多感測器夾爪之開發與其在多外形與多尺寸物件自動化取放任務之應用

Abstract

隨著科技的發展與產業結構的改變，產線自動化已成為製造業重要的發展方向，而其中物件的自動化夾取與放置，不僅是產線上物料移動與組裝任務中不可缺的一環，未來也更是進行如雙手裝配任務之類的複雜自動化操作的必要動作之一。本研究承襲國內自動化發展趨勢，開發新式具多重感測與智能的機械夾爪，並以此夾爪針對自動取放任務設計一系列演算法則。夾爪具有仿人類抓取的三指和兩指雙重使用模式，每一指上並具有被動適應性設計，綜合此兩項機械特性可以讓夾爪產生多種不同夾取姿態以適應各種物件。夾爪上亦配置壓力陣列、電位計、力感測器、與加速度規等共四種感測器，產生進行智能化夾取所需的豐富感測回饋資訊。除此之外，為能以實驗方式測試夾爪特性，本研究並設計製作SCARA(Selective Compliance Assembly Robot Arm)機械手臂，並為其規劃平穩運行軌跡，以提供夾爪在空間中四個自由度的運動。完成整體系統架設後，開發自動取放之演算法，使該系統能於既定空間中，找出物件位置並判別出物體形狀，再以夾爪最適合姿態進行穩定的物件夾取。物件位置搜尋主要依賴觸覺感測和加速度資訊，物件外形識別與尺寸估算則倚賴觸覺感測陣列與夾爪姿態資訊，從而分辨出不同尺寸之圓柱，圓球，正方體、以及長方體等物件。穩定物件抓取則同步利用感測器回饋與夾爪可變外形特性，來形成力控制穩定抓取或幾何封閉包覆夾取。前者配合演算機制能以接近最小的抓握力道夾取未知重量的物體，並在物體受到外在干擾時能自我調整夾取力道，以防止物體滑落。而後者則利用夾爪特有的被動元件適應特性，使夾爪包覆物體並限制其運動，達到穩定夾取的功能。總整上述各部分開發成果，以達到完整多尺寸多外形物件的自動取放任務。

With the change of industrial structure and the development of technology, automated production line has been the main role in manufacturing. Among automated production line techniques, automatic pick and place operation is an indispensable part in material transfer and assembly. It is also necessary for advanced applications such as dual arm assemblies. Following the trend of progress in automation, we develop a robotic hand with sensory capabilities and design a series of procedures for pick and place operation. Our robotic hand has three fingers with underactuation. Two of them are rotatable, thus it can mimic human grasping behaviors such as cylindrical grasp and spherical grasp. Utilizing various grasping patterns, our robot hand can grasp a wide range of objects. Furthermore, with sensor-rich feedback control involving instruments such as pressure arrays, potentiometers, FSR sensors and an accelerometer, our robotic hand has the ability to determine and adjust for different grasping scenarios. In order to test the robotic hand’s capabilities, we also design a SCARA (Selective Compliance Assembly Robot Arm) robotic arm and add trajectory planning functionality for smooth object manipulation. After finishing all the hardware, we create an automatic pick and place algorithm enabling the robotic hand to find the position of an object, recognize its geometry and select the most appropriate grasping posture for stable manipulation. FSR sensors and the accelerometer are crucial for performing the two main actions: object searching and geometric recognition. These two actions rely on pressure array sensing and the kinematics between fingers. The robotic hand can figure out different basic shapes such as cylinders, spheres, cubes and rectangular solids of different sizes. As for stable manipulation, the robotic hand implement force-closure and form-closure gripping. The former gripping pattern allows the robotic hand to grasp an object with minimal force and can also adjust grip force while receiving disturbance. The latter grasping pattern uses its passive ability to encompass objects and resist motion. To summarize the project, the robotic hand can automatically grasp objects of different sizes and shapes, which is effective for Pick and Place operation.