Humanoid Robot Intelligent Walking Control
           

- 指導教授 黃漢邦 博士 研究生 程登湖

- Advisor :Dr.Han-Pang Huang Student :Teng-Hu Chen

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

Abstract:

To achieve stable walking by an intelligent humanoid robot, this thesis proposes a center of gravity (COG) trajectory planner, immediate change footstep controller, ankle stabilizer, and stereo vision assisted footstep planner. After assembling these elements, the humanoid robot walked stably on rugged terrain and explored a road suitable for walking. Finally, to reduce energy and weight, we designed a leg mechanism with a larger weight to mass ratio for use on our strictly weight-limited humanoid robot.


For COG trajectory generation, we used a preview control COG generator for intelligent walking. Due to the difficulties after a sudden change of footsteps, we proposed an immediate modification of the foot placement controller. Combining the advantages of the two controllers, we planned the footsteps and COG trajectory dynamically and the robot walked more freely. The ankle stabilizer was implemented to reinforce walking stability. Utilizing the controllers, our robot walked stably on the most rugged terrain. Finally, stereo vision is the window into the core of our humanoid robot. We made our humanoid robot more intelligent by composing stereo vision data and dynamic footstep planning. With this intelligence, the robot could analyze, all by itself, its environment and find the locations of obstacles, accessible regions, and the route to its destination. Finally, in order to reduce the weight and energy consumption of the humanoid robot, we used Finite Element stress analysis to remove any unnecessary materials and, on the other hand, to keep the structure strong enough to maximize the payload.


Finally, our simulation physical environment was constructed on ADAMS, and all of the control code was built in MATLAB. These two environments are connected by Simulink in MATLAB. Stress analysis was done using COSMOSWorks in SolidWorks.





中文摘要:


為了達成智慧型機器人的穩定行走,本論文的目的在於建立機器人全身重心軌跡規劃器、即時重心軌跡與腳步改變控制器、腳踝穩定控制器、視覺輔助腳步軌跡規劃 器,由這些控制器的組合,人型機器人才能穩定地行走在崎嶇的路面上,智慧地判斷並規劃出適合自己行走的路徑。最後,為了減少移動能量與重量,我們設計出較 大荷重與質量比值的腳部機構來安裝在重量被嚴格受限的人型機器人上。


在重心軌跡規劃生成方面,我們使用Preview Control 來生成重心軌跡,但有鑑於此控制方法,有著不易即時改變步態的缺點,所以我們提出了腳步即時改變控制器,結合此兩種控制器的優點,可使機器人達到步態與重 心的即時規劃,以及更高的行走自由度。而腳踝穩定控制器是用來強化機器人的行走穩定性,藉由此控制器,可以穩定地行走於大部分的小崎嶇路面。視覺是機器人 的智慧之窗,所以我們靠著機器人立體視覺的寶貴資訊加上視覺輔助行走軌跡規劃,賦予機器人智慧,這種智慧可以讓機器人分析自己身處的環境,並進一步分析出 障礙物位置分佈與可行走的路徑,憑藉著兩者的結合,機器人能獨立地行走至目的地。最後,為了減少移動能量與重量,我們利用有限元素應力分析,來去除多餘的 材料,並保持結構的強度,設計出較 大荷重與質量比值的腳部機構。


我們模擬的物理環境均建立在ADAMS上,而所有的控制程式均在MATLAB上撰寫,兩者透過MATLAB的Simulink做連結。而機構的應力分析應用SolidWorks之CosmosWorks軟體做驗證。