機器人主要目的就是能夠服務人類,負責從事各種的工作任務。而過去學者們因應不同情況或不同任務提出新方法或是架構,希望能滿足人類的需求。對於實行機械手臂抓取物體時,有些研究者利用控制理論來控制機械手臂或者根據影像特徵來實現抓取,前者會導致計算複雜度過高,後者雖然能夠分類出可以穩定抓取的區域或者是點,但是並沒有考量到額外的因素,使得任務失敗。為了解決上述情況,利用環境心理學提出的affordance概念應用在本研究系統的機械手臂抓取行為中,藉由此概念把感知、物體與任務知識以及行為結果之間的關係作連結,並由規劃器產生出一連串的動作,且利用OpenRAVE作為模擬環境執行,能夠展現出機械手臂對於環境的了解。在劇本中呈現出任務執行中有affordance的情況下,會比沒有affordance時完成任務的效率來的好。 Robots are responsible for carrying out various tasks to provide service to human. Researchers usually attempt to use new methods or new architecture for robot control in different situations or tasks to satisfy human needs. When implementing robot grasping, some researchers use control theory or make use of many sensory input to control robot arms to grasp objects. These approaches have introduced major problems like high complexity of computing or failed tasks relying solely on image features to classify stable area and unstable area for robot grasping. Therefore, we use affordance for robot grasping in our system. By linking the relation between perception, the knowledge of objects and task-action effects. The system produces a series of actions through the planner and uses OpenRAVE as simulation environment. The experiments show that the task execution with affordance performs more efficiently than without affordance.