In this research we develop a knowledge engine to give robots better understanding of human commands. The architecture uses ontology to determine how to hold an object in which orientation and makes robot understand the natural language command with spatial intelligence. In our approach, at first, the input command given in natural language is parsed using IBM Natural Language Understanding service, which identifies the action, subject and object. Then the system matches action and pattern with the list of skills available for the robot. The robot has a model database and a list of objects remaining in front from its sensors. From that list it will able to find the subject and object positions. From the model database it can find the object type and from the ontology the robot can check whether the subject can be put on top of the object. If it is not valid then robot will inform the user that the command cannot be performed. If it is valid then, the robot will perform the put operation.