A semantic knowledge-based method for home service robot to grasp an object

To help the home service robot grasp an object with the constraints from the object itself, task and environment at the same time, we propose a knowledge-based method. Firstly, we construct five ontologies (object, task, environment, constraint and probabilistic ontology) to represent the five in a unified format. The remarkable things are that the object ontology is part-based and the constraints (grasp type, grasp location, approach direction, opening width, trajectory constraint and grasp force) are throughout the whole grasp process. To reason the constraints out with the findings, we propose a collaborative reasoning mechanism. It contains ontological reasoning, rule-based reasoning and probabilistic reasoning. Secondly, we propose to use SWRL (Semantic Web Rule Language) rules to label the object’s segments with functional parts. For its textual form, this method is fast. Thirdly, JSHOP2 planner is used to make the inferred constraints as the parameters of grasp subactions. Once the subaction list is output, the corresponding executable codes are called. Finally, we use Tiago robot to grasp different objects in different task and environment in our laboratory.