標題: | 機器人順應性工作之規劃, 控制, 與協調 Planning, Control, and Coordination in Operating and Teleoperating Robot Compliance Tasks |
作者: | 官振鵬 Kuan, Cheng-Peng 楊谷洋 Young, Kuu-Young 電控工程研究所 |
關鍵字: | 機器人順應性工作;加強式學習;順滑模式控制;阻泥控制;遠端操控;虛擬實境;智慧分配協調;Robot compliance tasks;Reinforcement learning;Sliding mode control;Impedance control;Teleoperation;Virtual reality;Coordination and distribution in intelligence |
公開日期: | 2000 |
摘要: | 為了完成機器人順應性工作,精密的工作規劃以及良好的控制是不可或缺的。然而,由於對順應性工作的特性理解不足,再加上執行工作時所遭遇的外在環境不確定性的影響,要得到正確的工作規劃相當不容易;此外,在執行順應性工作時,位置與力量需同時加以控制,與環境也會有不可避免的碰撞,而正確的碰撞模式是非常困難建立的,也因此增加了控制上的複雜度。本論文針對順應性工作提出一控制架構,其中在近端與遠端均賦予智慧控制能力,彼此以合作與協調的方式進行工作規劃及控制,以操作機器人完成工作。我們將探討近端與遠端的規劃與操作如何達成有效的分工,其中提出了加強式學習機制來達成工作規劃,也發展了數個順應性控制策略用以進行順應性工作的執行。此外,為了進一步探討近端與遠端之間的合作關係,我們以人類操作者擔任近端的智慧提供者,並配合虛擬實境技術,發展一機器人遠端操控系統,據此,我們討論如何根據順應性工作的特性和要求,將智慧適當地分配給人類操作者與智慧型遠端控制器。我們以模擬與實驗的方式進行多種順應性工作,以展示所發展系統的能力,並分析在面對不同工作時,因應合宜的規劃、控制、與協調。 When robot manipulators are used to perform compliance tasks, exquisite planning and control are necessary. It is quite difficult to achieve accurate task planning due to insufficient information about the tasks and the uncertainties occurring during task execution. The complexity in control of compliance tasks also arises from simultaneous control of both position and force and inevitable contact with environments. And, the modeling of the dynamic interactions between the robot and the environment is complicated. In this dissertation, we propose a control structure for robot compliance tasks, in which intelligence and control load are distributed between the operator and remote sites in performing task planning and control. We focus on the issues of planning, control, and coordination between the operator and remote sites. A reinforcement learning mechanism for task planning and several compliance control schemes for task execution are then developed. Moreover, for further investigation on the cooperative relation between the operator and remote sites, we developed a telerobotic system. We are interested in how the intelligence in control to be distributed between the human operator and the remote intelligence controller according to the specific characteristics of a given compliance task. The telerobotic system for compliance tasks is developed using the VR(Virtual Reality) technique. Simulations and experiments based on various types of compliance tasks are performed to demonstrate the effectiveness of the proposed system, and to analyze the issues of planning, control, and coordination between the operator and remote sites in facing different tasks. |
URI: | http://140.113.39.130/cdrfb3/record/nctu/#NT890591002 http://hdl.handle.net/11536/67768 |
Appears in Collections: | Thesis |