9408545 Iwasaki When humans reason about the behavior of the physical world, reasoning with images plays a crucial role. However, artificial intelligence has so far mostly ignored visual reasoning, concentrating instead on the use of propositional representation methods, which are not necessarily suited for representing primarily spatial information. The proposed work aims at developing techniques for making use of diagrammatic reasoning techniques in the context of predicting behavior of spatially interacting objects. A general architecture for diagrammatic reasoning systems will be designed and programs based on architecture that solve such problems using a diagram in a way similar to that of humans will be implemented.
