3D Object Recognition and Pose Estimation
           

- 指導教授 黃漢邦 博士 研究生 賴威志

- Advisor :Dr.Han-Pang Huang Student : Wei-Zh Lai

Lab. of Robotics., Department of Mechanical Engineering, National Taiwan University, Taiwan

Abstract:

As the gradually aging population and the rise of automatic production, robots become necessary to us. Many types of robots have been proposed, the humanoid robot, by its human like appearance and geometry is an active field in robotics. However, the robot balancing problem is still a tough question. Biped attracted many researchers to develop the control rule and strategy. Because of the improvement of technology and the accumulated experiences, researchers in America and Japan start to develop their new robots which can run and jump smoothly and stably. The bipeds nowadays start to be able to deal with these difficult tasks. The first problem we met is to keep robot balancing while absorbing disturbances from the environment. In this thesis, we proposed a stabilizer with a real-time feedback control system to deal with balancing problem. We developed an optimal ZMP distributor as the control references. The COG Jacobian is used for whole body coordinate balancing. In addition, we used force sensors on ankles and Inertia Moment Unit (IMU) on waist. We did three experiments for proving the workability of this stabilizer. With the proposed stabilizer, the biped can endure external forces and walk on uneven terrain. The communication system of this robot is based on CAN-Bus and EtherCAT. It is a real-time system.





中文摘要:


機器人領域的發展除了材料與電腦的進步之外,仍需要更多經驗的累積。直到近幾年,在日本與美國尖端研究者篳路藍縷的開拓下,機器人領域厚重的大門才被緩緩推開,開始出現一道曙光。而首先面臨的問題,即是如何使機器人適應不同環境,以及在外界干擾下如何使機器人持續運作。本論文提出一套穩定性系統可使機器人在外界干擾或是環境發生變化時維持平衡。此方法結合機器人領域中最主要的兩種控制架構-位置控制 (Position Control)與力量控制 (Force Control),配合即時回饋控制 (Real-Time Feedback Control),改變機器人姿態以符合滿足預觀控制 (Preview Control)下的零力矩點 (Zero Moment Point)軌跡。此外,透過六軸力規與慣性量測儀(IMU)並導入質心Jacobian (COG Jacobian)之概念,設計一及時穩定器。本論文利用質心水平移動來補償線動量 (Linear Momentum Compensation),在角動量 (Angular Momentum Compensation)方面則是藉由上半身擺動來補償零力矩點誤差。利用更改逆運動學 (Inverse Kinematics)欲追蹤之目標座標,擺動腳 (Swing Leg)可以智慧地適應環境。如此一來,當機器人面對不同的地形以及無法預測之外力時,可智慧地調整其各軸軌跡,使行走之適應力明顯提高。本論文理論實現皆透過實驗室自主開發之人型機器人完成。此機器人透過USB-CAN Bus 以及 EtherCAT溝通,是一即時控制系統。