KNOWLEDGE REPRESENTATION
The main parts of the graph based knowledge representation are the knowledge base, the hypothesis storage, and one or
more dictionaries.
A. Knowledge base The knowledge base holds the knowledge of the system.
It is realized as a graph-based structure, divided to abstract and instance domains. In the abstract domain, the nodes denote the known abstract objects and concepts and the directed edges between them describe their relationship,
thus describing the knowledge of the system about the world in general.
In the instance domain the nodes represent existing objects that the intelligent machine (robot, iSpace) knows about.
The nodes are homogeneous and nameless (for an example, see Fig. 1).
Whenever the world of the robot is altered in any way, the changes should appear in the appropriate domains of the
knowledge base as well (physical changes (addition or removal of sensors, detectors, agents, etc.) in the instance domain,
while property changes (temperature of the room, etc.) in the abstract domain).
Figs. 2 and 3 show structural examples for the realization of the knowledge base. (All the examples are taken from the
ISpace application published in [10].) Fig. 2 depicts that tea, coffee, and cappuccino are drinks, beverages are synonymousto drinks, drink machines can make drinks, brewing is (in this case) synonymous to making, and actions done to drinks are heuristically dependent on time and mood. Fig. 3 shows an