In this thesis the semantics of grasping get developed first conceptually, then modelled ontologically and finally implemented in fortiss’ Robot Instruction Framework. The ad- vantage of doing this on a symbolic level compared to classical grasp planning approaches is that it saves a lot of computational effort and allows semantically meaningful reasoning about grasp modes. By logically guaranteeing incompabilities, the remaining search space for subsymbolic methods shrinks considerably. The basic idea is to compare the grasping capabilities of a gripper with the geometric conditions of the object to grasp, in order to find matching candidates - the applicable grasp modes. Basic sanity checks like values within ranges are performed and feasiblity scores per grasp mode allow ranking the re- sulting list of matches.
«
In this thesis the semantics of grasping get developed first conceptually, then modelled ontologically and finally implemented in fortiss’ Robot Instruction Framework. The ad- vantage of doing this on a symbolic level compared to classical grasp planning approaches is that it saves a lot of computational effort and allows semantically meaningful reasoning about grasp modes. By logically guaranteeing incompabilities, the remaining search space for subsymbolic methods shrinks considerably. The bas...
»