自主型四足機器人跨越河石地形的路徑規劃與實驗

Abstract

由於仿生四足機器人應用的範疇包括如軍事、工業、娛樂等方面，十 分廣泛，伴隨而來的便是越來越複雜的地形與環境適應力等技術需求，才 能使四足機器人真正符合現今社會的應用。有關四足機器人的研究發展方 面，目前有許多研究針對平坦路面、障礙跨越等路徑規劃問題進行探討， 本實驗室曾開發直線與迴轉最佳步伐的法則。本研究擬針對多種路徑規劃 方法，包括路線圖、單元分解法、勢場法及PRM(Probabilistic-Roadmap)等 進行進一步評估與推導，擬由上述方法中，分別擷取個別優點後建立一穩 定且有效的路徑規劃模式，應用至非連續地形的路徑規劃問題上，完成機 器人在特徵空間上之路徑之規劃，並利用代克斯托演算法(Dijkstra's algorithm)來尋找路徑中之最短路徑，減少機器人由初始位置移動至目標位 置的時間。本研究規劃以微處理單晶片(如：BASIC Stamp 2 或單晶片8051 等)組合而成機器人之控制系統，控制機器人之動作進行實驗來驗證前述路 徑規劃的可行性，最後進行誤差分析並提出改進之方式。 本研究第一階段主要目標為設計與規劃自主型仿生四足機器人在非連 續地形(本研究稱之為河石地形) 上跨進時之路徑規劃並透過動態模擬進行 驗證。第二階段將設計資料具體實現為實作成品，加裝微處理器晶片組， 並加入穩定性的監測感測器，經由感測器回饋機器人在所規劃路徑上的運 動情形與位置誤差後，配合所發展的路徑規劃系統，即時補償控制機器人 以避免失步與傾倒。第三階段則考慮起伏型與斜面型河石地形的步伐規 劃，由於須考慮四足機器人傾斜時，機體重心將嚴重影響機構姿態與動力 學問題的求解，本研究將結合力學平衡的法則推導與步伐規劃的控制，發 展適合於一般河石地形等非連續路面的控制法則，並以實作的四足機器人 系統加以驗證。

The quadruped robots are widely applied in various occasions, such as military , industry , entertainment etc. Many basic studies about the moving gait planning, static and dynamic stability, control algorithm, sensor fusing technology, etc. have been discussed. The optimal and efficient gait planning algorithm on both straight and turn around moving situation have also been investigated and proposed by NCTU CIDM Lab. this year. The study focus on the innovative development of single chip autonomous quadruped robot with a dextrous gait controller and skillful gait planner adaptive to specified roadmap especially for a non-continuous terrain (in the study is called river stone terrain). The river stone terrain is quite common existing on the living environment such as shortcut across a wild river, side walk with debris, building area with materials lie scatter, or even cliff climbing situation. The gait strategy for river stone terrain will be driven to the challenge of limit footprint region and highly stability concern. The indeterminacy roadmap beyond the sensible region is also one of the critical assignment need to be overcome. The current study is proposed to synthesize an efficient and valid method based on some critical issues on efficiency, stability, mechanical balance, probability roadmap generation in configuration space etc. The project will be implemented in three steps. First, the study is aimed at the gait planning methodology development for a quadruped robot scrawl on the special river stone terrain. In the proposed research project, the robot control system is currently evaluated to adopting the micro chip of BASIC Stamps , 8051 or others for further applications based on the PRM planning algorithm. Second, the control algorithm and practical mechanism with control-sensor unit will be designed and constructed to be a demonstrated model for verification experiments. Third, the developed gait planning model will be extending study to a model of inclined terrain even with an cliff topology. The results of proposed research will develop an efficient and valuable methodology for quadruped robot especially to be applied to a river stone terrain. The innovative contribution will also support the design of transportation vehicle applying in a river stone terrain.