人型機器人之全身控制與規劃

Whole Body Control and Planning for Humanoid Robots

李泓逸(Hung-Yi Lee)

指導教授 ： 黃漢邦

國立臺灣大學/工學院/機械工程學研究所/碩士(2012年)

https://doi.org/10.6342/NTU.2012.00948

摘要

在這篇論文裡，我們主要在試著控制我們實驗室所研發的整隻機器人。我們的目標是藉由控制機器人的重心和手腳末端點的位置來完成一些任務。
我們使用preview control來生出重心軌跡，接著把重心軌跡和末端點軌跡送進IK裡面做運算，運算過後就可以得到每一軸的軌跡。我們的IK運算了24個軸，其中手跟腳都各有六個軸。使用裝在機器人手腕和腳踝的六軸力規，機器人可以用三種不同的方式來移動物體。他可以使用推車推重物，用雙手抬起物體，或是使用一隻手提物體並且讓另外一隻手做其他的任務。在手腕的六軸力規可以量到物體的重量或是施於推車的力量。在腳踝的力規可以量到腳底板的反作用力，並且以此計算機器人的零力矩點。機器人可以利用這些資訊來調整重心的軌跡。
雖然我們一開始就讓機器人知道物體的位置和形狀，但機器人在一開始並不知道物體的重量。如果機器人可以辨識出物體的種類的話，他就可以自己選擇適當的方式移動物體，如此一來就更有機會融入我們的生活裡。

關鍵字

人形機器人 ； 即時步態生成 ； 零力矩點 ； 全身運動規劃 ； 逆向運動學

並列摘要

In this thesis, we attempt to control the entire body of a humanoid robot that was developed by our laboratory. Our goal is to allow the humanoid robot to use its body to perform some tasks that are controlled by the Center of Gravity (COG) trajectory and the trajectory of the robot legs and arms.
Our study uses preview control to generate the robot’s COG trajectory and then the COG and end-effector trajectory are sent to the IK-solver to determine the trajectory of all of the joints. The IK-solver makes calculations for 24 joints, which include the six joints in each of the robot legs and arms. Using force sensors that was installed on the wrists and ankles, the robot can move objects in three different ways. It can push a cart to move a heavy object, use both arms to lift an object, and have one arm holding an object, while the other performs another task. The force sensors on the robot wrists can also measure the weight of the object or the force that is exerted on the cart. Similarly, the force sensor on its ankles can measure the reaction force of the footpad and calculate the Zero Moment Point (ZMP) of the robot. Based on this information, the robot is able to adjust the COG trajectory.
Although we have provided the robot with information about the shape and position of the object, the robot does not initially know the object’s weight. If the robot is able to recognize the shape of the object, it can choose a suitable method for moving it. This will allow the robot operate more easily in daily lift.

並列關鍵字

Humanoid Robot ； Real-time Walking Pattern Generation ； ZMP ； Whole Robot Motion Planning ； Inverse Kinematics

參考文獻

[5] T. F. Chan, and R. V. Dubey, “A Weighted Least-Norm Solution Based Scheme for Avoiding Joint Limits for Redundant Joint Manipulators,” IEEE Transaction On Robotics and Automation, Vol. 11, No. 2, pp. 286-292, 1995.

[7] T. H. Cheng, "Humanoid Robot Intelligent Walking Control," Master Thesis, Department of Mechanical Engineering, National Taiwan University, 2009.

[10] M. Griffis and J. Duffy, “A Forward Displacement Analysis of a Class of Stewart Platforms,” Journal of Robotic Systems, Vol. 6, pp. 703-720, 1989.

[11] K. Harada, S. Kajita, F. Kanehiro, K. Fujiwara, K. Kaneko, K. Yokoi, and H. Hirukawa, “Real-Time Planning of Humanoid Robot’s Gait for Force Controlled Manipulation”, IEEE/ASME Transactions on Mechatronics, Vol. 12, No. 1, pp. 53-62, February 2007.

[13] K. Harada, S. Kajita, K. Kaneko, and H. Hirukawa, “Dynamics and Balance of a Humanoid Robot During Manipulation Tasks,” Proceedings of IEEE Transaction on Robotics, Vol. 22, No. 3, pp. 568-575, 2006.