本研究規劃在大範圍環境中實現機器人同時定位與建圖(SLAM),研究議題包括階層式同時定位與建圖、地圖管理、與路徑閉合等。階層式同時定位與建圖的概念是將小範圍地圖存成區域地圖,再結合成為大範圍的環境地圖;地圖管理則是規劃有效率的地圖分層機制與地標管理程序;在路徑閉合議題探討機器人重回到走過路徑時,區域地圖建立之後的新地標可能會比對到先前地圖的舊地標,這些比對成功的舊地標訊息可以用來修正狀態估測的誤差。本論文所發展的階層式SLAM系統,可以提高在大範圍環境中SLAM任務的擷圖速度與降低電腦運算時間,並成功呈現路徑閉合的效果。 This thesis presents an efficient algorithm for robot simultaneous localization and mapping (SLAM) in large-area environments. Research topics include hierarchical SLAM, map management, and loop closure. The concept of hierarchical SLAM is to construct many small-scale local maps, and then combine all the local maps to form a large-area environmental map. This study also develops an efficient map management method which consists of the map partition mechanism and the landmark manipulation procedure. In the case of loop closure, new landmarks in present local map might be successfully matched with the landmarks in previous local map. The information of the matched landmarks can be utilized to improve the accuracy of robot state estimation. The experimental results show that the developed hierarchical SLAM system could improve the sampling time and reduce the computational time.