俱直覺性教導及反應重現功能之反應向量產生器於七自由度冗餘機器手臂之應用

Abstract

隨著科技的進步，離機器人融入我們日常生活的日子越來越近。屆時，人類與機器人將頻繁地在同個工作空間內，合作處理一些服務性、工業性任務，例如在工廠裝配線，實驗室任務，或是居家雜事。然而，隨著人與機器人之間的距離縮短，更多安全性的議題必須被注意討論。另一方面，機器人為因應未事前訓練的工作，快速的重新教導機器人學習一項新任務也人機合作中的困難的挑戰之一。因此，在此篇論文中，我們提出了一個俱有直覺性教學和反應性重現功能的7自由度冗餘機械手臂系統，並詳細的討論了球形-迴轉-球形（S-R-S）組態機械手臂的結構。 首先，在直覺性教學的階段，多模式的直覺性教導是利用解耦控制實現，並將教導的點設置為反應性重現階段的目標點。其次，在反應性重現階段，反應向量產生器（RVG）將分別對於障礙物以及目標點，產生手臂末端點、手臂肘，以及手臂身的排斥、吸引向量。此外，基於向量的線上軌跡產生器（vOTG）被設計來平滑RVG所產生的離散命令，以確保機械手臂運行的穩定性。因此，機械手臂可以反覆地重現在直覺性教學的階段所學的任務，並在重現的過程中，達到全手肢臂的避障。最後的實驗結果，是使用了國立臺灣大學智慧機器人及自動化國際研究中心(NTU- iCeiRA)設計製作的7自由冗餘機械手臂，以及Kinect深度感應器所開發的。

With the advancement of technology, robots will gradually come into our daily life. In the scenario of human-robot collaboration (HRC), robots would most likely share the same workspace with human beings when dealing with service and industrial tasks, such as tasks in assembly line, laboratory, or home environment. However, it cannot be denied that the closer robot and human are, the more safety issues would rise. On the other hand, robots sometimes need to be re-programmed on-the-task to tackle untrained works, which is one of the major challenge in this topic. Thus, in this thesis, we propose an intu-itive-teaching and reactive-replaying system for a 7-DoF redundant robot manipulator, and elucidate the case of spherical–revolute- spherical (S-R-S) type manipulator. To begin with, in intuitive-teaching phase, the multimodal intuitive teaching is achieved by decoupling control scheme, and the taught waypoints are set as target points (Goal) of replaying phase. Secondly, in reactive-replaying phase, the reaction vector generator (RVG) is advised to generate repulsive and attractive vector for obsta-cle avoidance and goal approach in dynamic environment. Furthermore, the vec-tor-based online trajectory generator (vOTG) is provided to smooth jerky commands from RVG. As a result, the robot can repeat tasks taught in teaching phase, and achieve active whole-arm collision avoidance between targets in replaying phase. Experimental results with NTU-iCeiRA 7-DoF arm developed in our lab and Kinect depth sensor are presented.