自主型四足機器人跨越非平坦河石地形之路徑規劃與動態模擬

Abstract

四足機器人在如今的社會中扮演越來越重要的角色，伴隨而來的是越來越複雜的地形條件，本研究旨在發展一套四足機器人行走於非平坦河石地形上的路徑規劃系統，搭配ZMP穩定性條件，並透過ADAMS模擬軟體的分析與實驗的驗證，使四足機器人能順利地安全跨越河石地形。 本研究中應用本CIDM實驗室先前開發的蒙地卡羅限制型特徵空間路徑規劃法，建立一在具有限制區域之特徵空間中的連續路徑規劃系統程式，來簡化多維度路徑規劃問題而轉變為數個二維問題的組合。此外在非平坦的河石地形中，每一塊河石具有高低起伏的落差，機器人配合地形必須規劃傾斜攀爬或上階的動作，接著以CIDM實驗室所開發之NC-F4四足機器人模型建立四足機器人跨越河石地形的路徑規劃程式，加上地形條件及ZMP穩定條件，即可將四足機器人跨越河石地形的每一步伐經由程式計算及規劃出來。 為了確認所規劃路徑的有效實用，本研究採用機構運動分析及模擬軟體ADAMS建構出四足機器人的模型，並輸入各驅動軸時序位移函數，觀察在跨步動作中可能發生的問題並加以修正，嘗試建立一個能夠符合機器人動作設定的模型，另外在本文範例中針對機器人跨步的一些影響參數進行分析與測試，提供參數化四足機器人之設計評估與經驗歸納數據，做為相關產業之參考。最後本研究以實驗的方式，建立一單晶片BH-3四足機器人控制系統，實際按所規劃的路徑，展現預期的運動效果，並設法加以檢測驗證。 經由非平坦河石地形的路徑規劃模擬範例及實驗的佐證，證實本論文中提出的規劃方法是簡單且有效的。利用此方法，四足機器人在面對如河石地形般高複雜度的不連續地形時，就能依照本論文中發展之程式所運算出的步伐持續前進，達成跨越之目標。

The applications of a quadruped robot are adopted quite extensive, such as military, industry, entertainment, etc. Higher and higher topographical and environmental adaptive capacity requirement pushes the research intensity focus on the autonomous, sensory technology and gait planning etc. The related research works about multi-leg locomotive have been afforded a big progress recently. There are quite a few research works focus on the gait development of discontinuous terrain such as the case of a quadruped robot climbing on an inclined wall surface only by holding the protuberance with planned gait control.

The study proposed an innovative path planning methodology to calculate and generate a gait for quadruped robot to walk through a river-stone terrain. The Monte Carlo searching method is proposed to generate an obstacle avoidance and shortest trajectory for foot print of each leg in the configuration workspace of the quadruped robot. And it was followed by ZMP stabilization checking rule to ensure that the robot is stable at any moment. To get with the terrain, the study also includes some typical motions task in the path generation function, such as walking aslope, climbing steps, etc. Then, every single gait data could be programmed by the above-mentioned method and softward system.

The programmed gait data have been transferred into ADAMS system to verify its correctness and stabilization of the planned gait for the quadruped robot walking through a discontinuous river-stone terrain. Some problems were observed and modified by the simulation results to match more practical situation. Finally, the study also build up an autonomous quadruped system NC-BH-3 to demonstrate the usefulness and efficiency of the gait generated by proposed methodology.