Walking Pattern Analysis and Control of a Humanoid Robot

Walking Pattern Analysis and Control of a Humanoid Robot Testing and Applications

- 指導教授黃漢邦博士研究生俞舒文

- Advisor :Dr.Han-Pang Huang Student :俞舒文

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

Abstract:
For the stability of the biped robots, the center of gravity (COG) motion is generated by a walking stabilization control that is based on the Zero-Moment Point (ZMP) trajectory. In order to achieve smooth walking pattern generation, ZMP trajectory planning is proposed. With simulation results, we can tell the COG velocity markedly decreases and the robot walks more stably with planning ZMP trajectory. Moreover, a correlation-based control (CBC) is developed to realize on-line COG trajectory planning. Through COG Jacobian, the CBC can generate whole body motion to adapt to the various environments. From the simulation results, we can conclude that our algorithms can efficiently enhance the stability of the humanoid robot.

中文摘要:

本文之主要目的為增強人型機器人在行走時的穩定性。透過結合ADAMS/Control和MATLAB/Simulink的動態模擬器，進而模擬和分析人型機器人的行走狀態以及其穩定性。並類比於人類步行的模式，分析影響機器人行走的各個因素，做一整合性的探討。

在機器人步伐行走方面，首先藉由分析機器人行走在不同環境下的步伐參數，以及透過零矩點(ZMP)的軌跡規劃來達到平滑的行走模式。模擬的結果顯示，適當的步伐參數以及零矩點規劃能大幅降低能量的耗損並增加機器人在行走時的穩定性。

在機器人行走的即時控制上，藉由模仿人類的運動模式，進而發展關聯性控制(correlation-based control)，透過所發展的控制方法可依據零矩點的回授訊號，即時改善人形機器人在行走時的穩定性。本文另發展一套RWLS（Robust Weighted Least-Squares）的演算法，以求得穩定的逆運動學（inverse kinematics）之解。最後透過動態模擬器的模擬結果加以驗證。