俱力量及順應性控制之擬人型雙手臂機器人應用於人類動作模仿

隨著科技的發展，智慧型服務機器人將漸漸地進入我們的日常生活中。為了能協助人們處理複雜工作，機器人必須具備多功能且靈活的操作技巧。尤其對人型機器人來說，以人的行為動作來完成任務將是不可或缺的能力。為了讓機器人有類似人類的動作，機器手臂的軌跡規劃是重要關鍵。然而，由於需要同時處理多個自由度，使得產生類似人類動作的軌跡非常的複雜。因此，本論文的動機與目的在於發展一個人類動作模仿系統來使機器人產生類似人類的動作。透過示範來學習是一個直觀和有效的方式，讓人型機器人學習各種人類的動作。利用動作擷取系統直接將人的動作資訊提供給機器人，而不去進行複雜的軌跡規劃。在本論文中，我們使用具深度量測之視覺感測器Kinect來獲得人體的動作資訊，並即時轉換成機器手臂的運動軌跡。考慮到機器手臂的性能及軌跡的平滑性，我們設計了一個對速度及加速度做限制的線上軌跡產生器以得到平滑的軌跡。在機器手臂控制的部分，我們採用基於笛卡爾空間的控制架構，求得機器手臂在空間中移動所需的力，並透過向量投影的方法將笛卡爾空間的力轉成關節空間的力矩，再以扭矩控制來操控機器手臂。我們利用虛擬彈簧阻尼元件來實現許多功能，包括運動跟隨控制、虛擬牆及自我防撞的功能。虛擬牆的概念可用於限制機器手臂的工作空間，自我防撞的功能可以避免可能發生的碰撞。由於機器手臂的行為是由虛擬彈簧阻尼元件產生的力所決定，為確保機器手臂運行時的穩定，我們對力的大小及變化做了限制。因為系統所有的功能整合的十分完善，我們成功地展示了雙手臂機器人實時模仿人類動作的功能，並能確保其穩定性與安全性。

關鍵字

人類動作模仿；動作擷取；線上軌跡產生器；自我防撞；人型機器人

並列摘要

With the advancement of technology, intelligent service robots will gradually join with human society and come into our daily life. In order to help people deal with complex tasks, the robot must have versatile and flexible manipulative skills. Especially for humanoid robots, it will be indispensable that robots can accomplish the tasks with human action. To allow the robot has human-like movements, trajectory planning of robot arm is crucial. However, due to the need to handle multiple degrees of freedom (DOFs) simultaneously such that the generating human-like motion trajectory is very complicated. Therefore, the motivation and purpose of this thesis is to develop a human motion imitation system to make the robot generate human-like motions. Learning by demonstration is an intuitive and efficient way to let a humanoid robot learn a variety of human motions. By using the motion capture system, we can provide the human motion information to the robot directly, rather than take more cumbersome way for performing complex trajectory planning. In this thesis, we use Kinect which is capable of obtaining visual 3D depth information to get the human motion information and instantly convert the data into robot arm trajectory. Taking into account the capability of robot arm and the smoothness of trajectory, we design an on-line trajectory generator imposing the limit of velocity and acceleration to obtain a smooth trajectory. In the aspect of robot arm control, we use Cartesian-based control architecture to compute the required Cartesian force for moving the robot arm in the space. Through the vector projection method, the Cartesian force can be transformed into joint torque. Then we apply torque control to manipulate the robot arm. We use virtual spring-damper elements to implement many functions, including motion following control, virtual wall and self-collision avoidance. The concept of virtual wall can be used to restrict the workspace of robot arm. Self-collision avoidance can avoid possible collisions. Since the behavior of the robot arm is according to the force generated by the virtual spring-damper elements, we limit the magnitude and variation of the force to ensure the action of robot arm is stable. Because all of the functions in the system are well integrated, we successfully demonstrate that the dual arm robot can imitate human motion in real time and can guarantee its stability and safety.

並列關鍵字

human motion imitation ； motion capture ； on-line trajectory generator ； self-collision avoidance ； humanoid robot

參考文獻

[1] F. Wang, C. Tang, Y. Ou, and Y. Xu, "A real-time human imitation system," World Congress on Intelligent Control and Automation (WCICA), 2012, pp. 3692-3697.

[2] V. V. Nguyen and J. H. Lee, "Full-body imitation of human motions with Kinect and heterogeneous kinematic structure of humanoid robot," IEEE/SICE International Symposium on System Integration (SII), 2012, pp. 93-98.

[4] T. Petric and L. Zlajpah, "Smooth transition between tasks on a kinematic control level: Application to self collision avoidance for two Kuka LWR robots," IEEE International Conference on Robotics and Biomimetics (ROBIO), 2011, pp. 162-167.

[5] A. Thobbi and S. Weihua, "Imitation learning of arm gestures in presence of missing data for humanoid robots," IEEE-RAS International Conference on Humanoid Robots (Humanoids), 2010, pp. 92-97.

[8] J. B. Cole, D. B. Grimes, and R. P. N. Rao, "Learning full-body motions from monocular vision: dynamic imitation in a humanoid robot," IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2007, pp. 240-246.

國際替代計量

俱力量及順應性控制之擬人型雙手臂機器人應用於人類動作模仿

全文下載

主題瀏覽