Natural Walking Pattern Generation and Control for Humanoid Robots with the Design of Toe and Heel Mechanism

人型機器人之腳趾腳跟機構設計與自然步態生成與控制

趙毓文(Kenneth Yi-Wen Chao) ； 紀孟谷(Meng-Ku Chi) ； 嚴舉樓(Jiu-Lou Yan) ； 黃漢邦(Han-Pang Huang)

《中國機械工程學刊》 34卷2期 (2013/04) Pp. 151-158

摘要

近年來穩定的人型機器人雙足步態之實現已經由許多研究者所達成。大部分的學者使用倒單擺來做為生成步行軌跡的模型。而在這個基礎下，探討如何生成更加自然且省力的步行軌跡便成為了研究的重點。在傳統的方法上，人型機器人的步行生成採用了非平滑的零力矩點軌跡，並限制固定的重心高度，雖可生成穩定的步態軌跡卻可能造成機器人步行上不自然的動作，而且會消耗掉更多的能量。為了解決這些問題並增進步行的效能，針對零力矩點軌跡以重心高度的調整，還有加入腳趾腳跟機構為可行的方法。本文提出了一個自然步態軌跡生成的方法，除了採用平滑的零力矩點軌跡以及可調變重心高度的步態生成方法之外，也加入了腳趾及腳跟觸地等自然動作。此步態生成的效能以及腳跟腳趾機構的驗證則藉由模擬與真人大小的人型機器人實驗來呈現。

並列摘要

In recent years, stable walking for humanoid robot is achieved by many researchers. Most of them used inverted pendulum model for walking pattern generation. After achieving this, researchers start to focus on how to generate a more natural and power saving walking pattern. In conventional methods, non-smooth ZMP trajectory and constant COG height trajectory are used on humanoid robots. These methods can generate stable walking pat-terns but they are less human-like and consume more power. To overcome these problems, researchers tuned their COG and ZMP trajectory to attempt to generate natural walking patterns. Toe and heel mechanisms are also installed on humanoid robots to improve the walking performance. In this paper, a natural walking pattern generator with smooth ZMP trajectory, varying COG height trajectory and natural motions with toe and heel mechanism is proposed. The performance of the proposed method and mechanism is verified with simulations and experiments.

並列關鍵字

ZMP ； Varying COG height ； Preview control ； Natural walking ； Toe-off and heel-contact motion